Bioengineering doi: 10.3390/bioengineering13101175

Authors: İsmail Baydili Burak Tasci Gülay Tasci Sengul Dogan Turker Tuncer

Artificial intelligence (AI) is increasingly used across the diagnostic pathway, from screening and disease detection to differential diagnosis, prognostic stratification, and treatment-response assessment. This narrative review examines predictive, generative, multi-modal, and agentic AI through a diagnostic-science lens and evaluates the evidence required for translation into clinical practice. Predictive models can identify complex patterns in images, physiological signals, laboratory data, and electronic health records, but high retrospective accuracy does not establish diagnostic utility. Clinical validity and utility depend on the intended use, target population, disease spectrum and prevalence, reference standard, operating threshold, calibration, and the consequences of false-positive and false-negative results. Generative AI may support problem representation, differential diagnosis, information synthesis, and documentation, yet fluent outputs can omit critical alternatives, amplify false premises, or convey unjustified certainty. Multimodal systems may better reflect clinical reasoning by integrating imaging, text, signals, pathology, and molecular data, but they must demonstrate incremental value over the best single-modality test and remain robust to missing data. Agentic systems can coordinate evidence retrieval, test selection, and sequential workflows, thereby increasing the need for bounded permissions, auditability, error recovery, and human escalation. Across all paradigms, a credible pathway to diagnostic implementation requires independent external and prospective validation, representative consecutive patients, subgroup analysis, workflow and human–AI evaluation, clinical impact studies, and post-deployment surveillance. The field should therefore be judged not by model capability alone, but by whether AI measurably improves diagnostic yield, timeliness, safety, equity, and patient-relevant outcomes in real care settings.

Bioengineering doi: 10.3390/bioengineering13101174

Authors: Harshit Mishra Farhan Adam Nachiket Makwana Pradyot Tiwari Subramani Kandasamy K. V. S. Hari

Atrial fibrillation (AFib) is the most prevalent sustained cardiac arrhythmia, yet most deep learning detectors are validated on a single cohort and few report uncertainty. We developed a compact 14M-parameter convolutional neural network for binary AFib classification from 10 s, 12-lead ECG at 500 Hz, trained exclusively on the HDXML dataset (67,432 ECGs; 9628 AFib) with on-the-fly augmentation for noise, drift and missing channels. Using one unchanged checkpoint, we evaluated six independently sourced public cohorts totaling over 1.25 million recordings, including MIMIC-IV critical care and two ambulatory Holter cohorts. All metrics are reported with 95% confidence intervals from a cluster bootstrap at the patient, recording or subject levels. ROC–AUC exceeded 0.95 in every cohort, from 0.960 (95% CI 0.956–0.963) on CODE-15 to 0.999 (0.998–0.999) on SPH, including 0.966 (0.965–0.967) on MIMIC-IV with 75.3% sensitivity and 97.7% specificity, 0.963 (0.929–0.987) on CPSC 2021 and 0.995 (0.988–0.999) on MIT-BIH. Performance was stable under progressive lead masking without retraining, and median inference took 110 ms per segment on a server-class CPU without a GPU. A compact network trained on one curated source can therefore generalize across hospitals, countries and recording hardware. This study is retrospective; prospective validation is required before clinical use.

Bioengineering doi: 10.3390/bioengineering13101173

Authors: Graziella Di Grezia Alfredo Clemente Stefano Dugheri Vincenzo Mattei Alessio Gili Emanuela Mari Fabrizio Urraro Nicola Mucci Antonio Baldassarre

The COVID-19 pandemic exposed persistent weaknesses in health-data interoperability, cybersecurity, provenance, cross-organizational governance, and continuity of digital care. This short communication presents an updated focused narrative synthesis and a five-layer governance framework for integrating blockchain and the Internet of Medical Things (IoMT) in patient-centered digital health. The literature was updated through 10 August 2026, with recent peer-reviewed evidence complemented by authoritative standards and European regulatory sources. The synthesis positions blockchain primarily as a distributed governance, provenance, and audit layer rather than a clinical database or stand-alone security solution, while IoMT provides heterogeneous sensing infrastructure for continuous and distributed care. The framework separates: (1) device and data acquisition; (2) network and edge orchestration, including 5G/SDN where clinically justified; (3) blockchain-based governance; (4) privacy-enhancing data, analytics, and AI; and (5) application and service delivery. Homomorphic encryption, zero-knowledge proofs, secure aggregation, differential privacy, and federated learning are treated as complementary controls with distinct threat models and computational trade-offs. The revised framework also incorporates the European Health Data Space, the EU AI Act and its medical-device interplay, and NIS2 as a regulatory overlay. Across six use-case domains, evidence is strongest for technical feasibility, provenance, and programmable access control; evidence for clinical effectiveness, cost-effectiveness, equity, sustainability, and large-scale deployment remains limited. The proposed model therefore emphasizes standards-based interoperability, minimization of on-chain personal data, endpoint trust, auditable human oversight, and prospective real-world validation.

Bioengineering doi: 10.3390/bioengineering13101172

Authors: Jianwei Lai Aiguo Song Xuemei Fan Guangyu Sun Yaxuan Di Ting Wu

Soft robotic systems have attracted increasing attention for hand rehabilitation because of their compliance, adaptability, and potential for comfortable human–robot interaction. However, the clinical translation of soft hand rehabilitation robots remains limited by insufficient integration of interactive training, user feedback, and practical wearability. This study presents a multimodal feedback-enhanced soft hand rehabilitation robot based on lattice-structured soft pneumatic actuators (LSPAs), together with a preliminary clinical evaluation. The system integrates a fabric-based soft glove, pneumatic actuation, fingertip force sensing, hand-motion tracking, and interactive virtual rehabilitation tasks to provide coordinated visual, auditory, force-related, and mirror-visual feedback. Material characterization was first conducted to compare three TPU materials with different mechanical properties, followed by system-level evaluation of usability, wearability, and actuator durability. A four-week robot-assisted rehabilitation intervention was then conducted in nine stroke survivors. Participants completed five one-hour sessions per week, including passive training and multimodal interactive training. Clinical outcomes were evaluated using the Brunnstrom motor recovery stage, Fugl–Meyer Assessment of the upper extremity (FMA-UE), National Institutes of Health Stroke Scale (NIHSS), and Activities of Daily Living (ADL) scale, while usability was assessed using the System Usability Scale (SUS). After four weeks, the mean Brunnstrom stage increased from 2.44±2.07 to 3.67±2.12, FMA-UE increased from 18.33±25.63 to 32.67±23.11, NIHSS decreased from 5.67±4.36 to 3.22±2.99, and ADL increased from 49.78±39.53 to 66.67±22.50. The mean SUS score was 74.4. TPU420D provided the best durability among the tested configurations, with more than 30 h of recorded use without failure during the experimental period. These findings support the feasibility and usability of the proposed system and provide preliminary evidence for its application in stroke hand rehabilitation. Because of the small sample size, heterogeneous post-stroke status, short intervention period, and absence of a control group, the clinical findings should be interpreted as preliminary rather than confirmatory evidence of therapeutic efficacy.

Bioengineering doi: 10.3390/bioengineering13101171

Authors: Cheng Liu Yang Su Yao Wei Yongxin Li Chengxi Bao Rui Qiao

Aspirin prevents preterm birth in some women with preeclampsia but not others, yet no longitudinal multi-indicator study has examined subtype differences in aspirin response. We aimed to identify subtypes most likely to benefit. In this retrospective cohort of 53,362 deliveries (4505 preeclamptic women), we applied a two-stage machine learning framework using 15 indicators (14 laboratory tests plus systolic blood pressure) from two-time windows: before 16 weeks and within 2 weeks before delivery. A long short-term memory autoencoder with K-means clustering identified subtypes, and causal forest estimated the average treatment effect of aspirin on preterm birth (<37 weeks) for each subtype, adjusted for confounders. Five stable subtypes were identified (silhouette coefficient 0.542; mean adjusted Rand index 0.903). One subtype (n = 345), characterised by mild liver enzyme elevation and coagulation abnormalities in late pregnancy, had the highest preterm birth rate (57.7%) and ICU admission rate (9.6%), and showed the largest aspirin-associated reduction in preterm birth (ATE −0.026, 95% CI: −0.032 to −0.021), consistent across 70/30 splits. Sensitivity analyses using 34 indicators confirmed similar effects (ATE −0.027, 95% CI: −0.031 to −0.023), while no benefit was observed in non-preeclamptic women. These findings suggest that aspirin prophylaxis may be associated with a greater reduction in preterm birth in this preeclampsia subtype, but further validation is required.

Bioengineering doi: 10.3390/bioengineering13101170

Authors: Riya Gandhi Nora Molten Selorm Dente Quint Stein Angel Artega Emmett Reid Narutoshi Hibino

The development of patient-specific vascular constructs is essential for advancing clinically effective 3D-bioprinted and tissue-engineered grafts. Optimizing cellular composition and promoting robust vascularization within engineered tissues is important for building an effective vascular construct. In cardiac and vascular patches in particular, insufficient microvascular development often leads to diffusion limitations, ischemia, and eventual necrosis, representing a major barrier to clinical translation. This review examines the challenges of designing constructs that are both anatomically functional and adequately vascularized. We discuss current strategies for tailoring cell ratios and enhancing vascular maturation, as well as emerging technologies in bioprinting, biomaterials, and computational design. Together, these innovations highlight future directions for creating grafts capable of functional integration and long-term survival in patients.

Bioengineering doi: 10.3390/bioengineering13101169

Authors: Hui Chang Xiaolong Wang Yutong Wu Qian Li Qingsen Gao Xianhu Liu Chenchen Ren

Radiation-induced vaginal injury is an underrecognized complication of cervical cancer radiotherapy and evolves through a pathological continuum involving epithelial depletion, oxidative stress, chronic inflammation, microvascular dysfunction, extracellular-matrix remodeling, fibrosis, loss of tissue compliance, and vaginal stenosis. Current management remains largely supportive and is constrained by short local residence, limited anatomical adaptability, variable adherence, and insufficient control of radiation-specific tissue damage. This Mini-Review examines polymeric biomaterials for the prevention and repair of radiation-induced vaginal injury from a pathobiology–material–function perspective. Three complementary strategies are discussed: local polymeric formulations and mucoadhesive hydrogels for mucosal protection and sustained delivery; personalized dilators, stents, and adaptive devices for maintaining vaginal patency; and regenerative scaffolds and tissue-engineered constructs for restoring epithelial, vascular, stromal, and smooth-muscle compartments. Conventional hyaluronic-acid formulations have progressed furthest clinically, whereas responsive hydrogels, shape-adaptive devices, extracellular-matrix-derived materials, and cell- or exosome-based scaffolds remain predominantly preclinical. Importantly, much of the evidence derives from acute irradiation models, non-irradiated reconstruction studies, or engineering prototypes and therefore does not yet demonstrate durable reversal of chronic fibrosis or stenosis. Future development should prioritize stage-specific multifunctional interventions, fractionated irradiation models, standardized structural and functional outcomes, and rigorous oncological safety assessment. Acellular hydrogels, cell-free hybrid scaffolds, and removable drug-eluting devices may offer the most practical near-term translational routes.

Bioengineering doi: 10.3390/bioengineering13101168

Authors: Mingze Lyu Abhirup Banerjee

Carotid intima-media thickness (CIMT) is an established marker of subclinical atherosclerosis, but its reliance on ultrasound equipment and trained operators limits its suitability for large-scale screening. Retinal fundus photography may offer a more scalable approach to vascular assessment, as it contains retinal vascular features associated with systemic cardiovascular health and can be acquired quickly and non-invasively. This study proposes a two-stage transfer learning framework for CIMT-based cardiovascular risk classification. An attention-enhanced EfficientNet-B4 is first pretrained on the Asia Pacific Tele-Ophthalmology Society 2019 Blindness Detection dataset and subsequently transferred to a weight-sharing Siamese architecture to classify the China Fundus Carotid Intima-Media Thickness dataset. Across ten independent initializations, the proposed model achieved an average macro-F1 score of 77.81% and an area under the receiver operating characteristic curve of 84.08% on the validation set and 78.29% and 85.22%, respectively, on the test set. It achieved overall higher point estimates for performance metrics than those previously reported for the Siamese ResNeXt baseline model with squeeze-and-excitation attention, and demonstrated a reduced disparity in class-wise performance. Grad-CAM analysis further revealed differences in the spatial activation patterns between models trained under different experimental configurations. Overall, these findings support the feasibility of fundus imaging as a scalable and non-invasive approach to CIMT-based cardiovascular risk assessment.

Bioengineering doi: 10.3390/bioengineering13101167

Authors: Soyun Im Chulho Kim Yu-Seop Kim

Stroke is a leading cause of death and long-term disability, and auxiliary methods for rapid patient identification are needed. Previous studies based on electrocardiography (ECG) have attempted stroke classification using quantitative measurements and future stroke risk prediction using raw signals. However, measurements do not preserve all waveform detail, and models using raw signals alone may not sufficiently learn quantitative characteristics from limited data. We therefore propose a hybrid fusion framework that integrates the raw 12-lead signal and the quantitative measurements of the same ECG record. It combines a signal-only model, a feature-only model, and an interaction early fusion model that couples the two inputs at the raw data and representation levels, averaging their logits for the final prediction. Evaluation used 4147 ECG records from 4020 patients in a single-institution retrospective cohort, with patient-level splits. The proposed model achieved an F1-score of 77.6%, an area under the receiver operating characteristic curve (AUC) of 0.832, and a Brier score of 0.169. Among the evaluated models, the proposed framework achieved the highest AUC and lowest Brier score. The AUC improved by 0.034 over the best unimodal model, and an AUC of 0.820 was maintained under a year-based temporal holdout. These results suggest that integrating complementary representations of the same ECG at multiple levels can improve stroke classification. The framework could potentially serve as an auxiliary decision-support tool for identifying patients who require further evaluation, using ECG characteristics statistically associated with stroke.

Bioengineering doi: 10.3390/bioengineering13101166

Authors: Marcella Silva Vieira Roberto Sousa Dias Helena Santiago Lima Maíra Paula de Sousa Cynthia Canêdo da Silva Sérgio Oliveira de Paula

Sulfate-reducing bacteria (SRB) play crucial roles in microbiologically induced corrosion (MIC), especially in anaerobic environments where the production of hydrogen sulfide (H2S) and biofilm formation accelerate metal degradation. Conventional treatments, such as THPS-based biocides, while effective, face limitations relating to the development of resistance and poor biofilm penetration. In this study, we evaluated the use of a non-specific bacteriophage cocktail as a biological control strategy against a mixed SRB culture (AP) biofilm and related corrosion, utilizing a pilot-scale system with AISI 1020 carbon steel coupons. Over 38 days, we monitored H2S production, optical density, and ATP concentrations, along with the surface characterization of steel coupons through scanning electron microscopy (SEM) and profilometry. Metagenomic sequencing revealed changes in microbial community composition, including reductions in sulfate consumers and shifts in dominant taxa. SEM and profilometric analyses showed disrupted and less aggressive biofilms on treated surfaces, often exhibiting filamentous structures. Diversity indices indicated the modulation of microbial richness and composition over time. These findings demonstrate the potential of bacteriophage cocktails to reshape complex microbial communities and reduce MIC, providing a sustainable alternative to chemical biocides in large-scale industrial settings.

Bioengineering doi: 10.3390/bioengineering13101165

Authors: Aurelio Trofè Alessandro Piras Michela Persiani Luca Breviglieri Alessandra Laffi Andrea Meoni Milena Raffi

Pulsed electromagnetic field (PEMF) increases the amplitude of muscle activity during physical exercise. This study investigated the effects of PEMF stimulation timing on neuromuscular activation in young sedentary males, during moderate-intensity constant-load cycling. We examined whether a PEMF applied at the beginning or during the middle of continuous exercise elicits differential effects on muscle activation. The participants performed a 30 min cycling task at 50% of VO2max in which PEMF stimulation was alternately activated or deactivated in 15 min bouts. Surface electromyography (EMG) was recorded from the right vastus medialis and biceps femoris, and blood lactate samples were collected at baseline and during exercise. Data were analyzed via linear mixed-effects modeling. Active PEMF stimulation significantly increased normalized EMG amplitude in the vastus medialis (p < 0.001) and biceps femoris (p = 0.001) compared to inactive conditions, with the greatest effect observed when stimulation was applied after 15 min of exercise. Blood lactate concentrations were significantly elevated during PEMF application relative to baseline and inactive stimulation (p < 0.001). The timing of PEMF stimulation influenced neuromuscular responses during exercise, highlighting its potential as a tool to modulate muscle performance in sedentary individuals.

Bioengineering doi: 10.3390/bioengineering13101164

Authors: Taariq Mohammed Giuliano Scarcelli Osamah Saeedi

The biomechanics of the vitreous humor undergo important changes in both normal aging and various pathologic conditions. While these changes are clinically visible, there are limited objective in vivo measurements of vitreous biomechanics. New imaging modalities have the potential to enable quantitative assessment of vitreous properties such as viscoelasticity in vivo, which could have relevance in assessing the risk of progression of vitreoretinal diseases. In this article, we discuss the current clinical approaches to vitreous biomechanics and review relevant new imaging methods with potential for clinical translation.

Bioengineering doi: 10.3390/bioengineering13101163

Authors: Chao Xin Zhixiong Chen Zeyu Huang Jiucun Wang Luojun Lin

Domain generalization (DG) for medical image segmentation is commonly approached by simulating domain shifts through deterministic image transformations or feature perturbations. However, such methods implicitly assume that unseen domains can be approximated by predefined appearance variations. In this work, we challenge this assumption and hypothesize that medical domain shifts are more fundamentally characterized as uncertainty in spectral distributions. While anatomical structures are largely preserved across institutions, scanners, and acquisition protocols, substantial variations arise in image appearance, texture, and artifacts, which are predominantly manifested in the spectral domain. To test this hypothesis, we propose Spectral-Distribution Uncertainty (SDU), a frequency-aware DG framework that explicitly models uncertainty in the Fourier amplitude distributions of intermediate representations. By decomposing Transformer features into amplitude and phase components, SDU disentangles domain-specific spectral characteristics from domain-invariant anatomical semantics. Instead of imposing handcrafted frequency perturbations, SDU estimates spectral-distribution uncertainty during training and injects stochastic variations into the amplitude space, synthesizing diverse yet anatomically consistent representations that emulate potential unseen domains. Consequently, SDU expands the support of spectral feature distributions, enabling the model to learn representations that are robust to uncertainty-induced spectral shifts. Extensive experiments on retinal fundus and prostate MRI benchmarks demonstrate that SDU consistently outperforms state-of-the-art DG methods. Beyond empirical gains, these findings provide evidence that medical domain shifts are more appropriately characterized as uncertainty in spectral feature distributions, offering a principled perspective for generalizable medical image segmentation.

Bioengineering doi: 10.3390/bioengineering13101162

Authors: Margherita Tumedei Ugo Graziani Alessandro Cipollina Renato Ori Adriano Piattelli Ugo Covani Lorenzo Enia Calogero Pitruzzella Francesca Mangione

Background: Primary stability (PS) is a crucial determinant of long-term dental implant success, particularly in compromised or low-density bone sites. While macrogeometry is known to influence initial mechanical anchorage, the specific contribution of thread dimensions across varying bone densities remains incompletely understood. This in vitro study evaluated the influence of two different implant macrogeometries characterized by distinct thread configurations on primary implant stability within standardized polyurethane foam models representing different bone qualities. Methods: Two implant designs featuring distinct thread profiles were tested: the Evolution KRI (larger threads, pitch 1.8–2.2 mm, depth 0.55–1.025 mm) and the Evolution PRI (smaller threads, pitch 1.0 mm, depth 0.5 mm), across three fixture diameters (4.0, 4.5, and 5.0 mm). Implants were inserted into three standardized polyurethane block configurations: Group I (25 PCF uniform block, simulating D2 dense trabecular bone), Group II (10 PCF trabecular base with a 3 mm 40 PCF cortical layer, simulating D3 bone), and Group III (12.5 PCF trabecular base with a 2 mm 40 PCF cortical layer, simulating D4 bone). Each experimental group included 10 implants (n = 10). Two RFA measurements were collected for each implant and averaged prior to statistical analysis. Group comparisons were performed using one-way ANOVA followed by Tukey’s post hoc test. Primary stability was quantitatively assessed using Resonance Frequency Analysis (RFA) to record Implant Stability Quotient (ISQ) scores. Results: Statistically significant differences in ISQ values were observed among the experimental groups (F = 9.909,p < 0.0001,R2 = 0.4785). Implants featuring larger threads (KRI) demonstrated consistently higher mean ISQ scores compared to smaller-thread implants (PRI) across all bone densities. In Group I (D2 model), KRI implants achieved a significantly higher mean ISQ than PRI implants (67.1 ± 1.91 vs. 65.1 ± 1.43; p = 0.0472). In Group II (D3 model with 3 mm cortical layer), this difference was most pronounced, with KRI reaching 68.5 ± 0.58 compared to 64.8 ± 2.15 for PRI (p < 0.0001). In Group III (D4 model with 2 mm cortical layer), KRI maintained a higher stability trend (68.0 ± 0.93 vs. 66.5 ± 1.42), though the inter-system difference did not reach statistical significance (p = 0.2401). Inter-group comparisons confirmed that the highest overall primary stability was obtained by large-thread implants in the presence of a 3 mm cortical layer. Conclusions: Thread macrogeometry significantly influences primary implant stability; the tested KRI macrogeometry demonstrated higher ISQ values than the PRI macrogeometry under specific polyurethane configurations, particularly in intermediate trabecular–cortical bone structures.

Bioengineering doi: 10.3390/bioengineering13101161

Authors: Kazeem B. Olanrewaju Ngozi D. Mbue

Early differentiation of neurodegenerative movement disorders remains challenging because of overlapping cognitive-behavioral profiles. The Montreal Cognitive Assessment (MoCA) is a widely used 30-point screening instrument; however, the extent to which MoCA subdomain scores alone can support automated multi-class differential diagnosis or score prediction requires rigorous methodological evaluation. In this study, we investigated supervised machine learning algorithms across two distinct tasks using clinical data from the National Institute of Neurological Disorders and Stroke (NINDS) Parkinson’s Disease Biomarkers Program (N=64): (1) multi-class classification across movement disorder diagnoses, including Parkinson’s disease (PD, n=27), No Neurological Diagnosis (Controls, n=14), Progressive Supranuclear Palsy (PSP, n=9), Essential Tremor (ET, n=8), Corticobasal Degeneration (CBD, n=1), Multiple System Atrophy (MSA, n=1), and Other (n=4); and (2) numerical prediction/reconstruction of the MoCA total score from its constitutive domain items, with external validation on 1499 records from the Parkinson’s Progression Markers Initiative (PPMI). Under stratified 5-fold cross-validation, multi-class diagnostic classification was severely limited: the highest macro-averaged F1-score achieved by any trained model was 0.362 (95% CI: 0.284–0.440; Logistic Regression), trailing a naive Constant majority-class baseline in classification accuracy (0.463 vs. 0.366). Bootstrap resampling yielded apparent increases in performance (Neural Network macro-F1: 0.558, 95% CI: 0.512–0.604; AdaBoost: 0.555); however, this reflects optimistic bias from pseudo-replicate sampling rather than genuine clinical generalizability. For continuous score modeling, Linear Regression and Stochastic Gradient Descent achieved near-perfect internal fit (R2=1.000 and 0.996, respectively), and retained near-perfect performance in the external PPMI cohort (R2=0.992 and 0.991; MSE ≤0.077). We demonstrate that this regression success reflects the mathematical recovery of an additive scoring definition, i.e., target leakage, rather than independent clinical prediction. MoCA domain scores alone possess insufficient disease-specific variance to differentiate complex parkinsonian syndromes, underscoring that automated clinical classification requires the integration of multimodal neuroimaging, fluid, and digital motor biomarkers.

Bioengineering doi: 10.3390/bioengineering13101160

Authors: Ahmad Hammoud Andrey Briko Vladislava Kapravchuk Artem Malakhov Sergey Shchukin

Evidence for pulsed electromagnetic fields (PEMFs) in peripheral circulatory dysfunction is heterogeneous, while limb bioimpedance may index segmental pulsatile hemodynamics. This scoping review mapped PEMF effects, bioimpedance-based vascular assessment, and adaptive stimulation architectures. Scopus, PubMed, and Web of Science were searched in July 2026, supplemented by IEEE Xplore during revision (PRISMA-ScR reporting; no registered protocol). The main search used AI-assisted sequential investigator review; IEEE records underwent independent duplicate human screening. The main search yielded 1750 unique records and 30 core reports; IEEE Xplore added two technical reports. The combined corpus comprised 32 reports, conservatively grouped into 30 study/report clusters. It contained 15 direct clinical, six clinical support/proxy, five preclinical/veterinary, and six technical/pilot reports; 26 were assessed in full text and six from abstracts. Reported outcomes included positive, mixed, and negative directions; heterogeneous endpoints and designs, and the absence of formal risk-of-bias assessment precluded efficacy inference. Low-frequency, low-millitesla regimens recurred, but incomplete and non-equivalent dose reporting precluded identification of an optimal regimen. Contextual bioimpedance evidence supported segmental impedance and pulse-related hemodynamic assessment, not direct measurement of capillary microcirculation. No validated system adapting PEMF from limb bioimpedance for a peripheral-circulation target was identified in the databases searched. The proposed adaptive architecture requires prospective technical and clinical validation.

Bioengineering doi: 10.3390/bioengineering13101159

Authors: Jiongxiang Zhao Enze Shao Zilong Lu Jun Wen Chao Zeng Tibor J. Goda Viktória Tafferner-Gulyás Dong Sun Yaodong Gu

Motorsport remains high risk despite advances in crash protection, vehicle design, and medical response. Effective injury prevention requires understanding interactions among crash loading, human injury tolerance, protective-system design, and material behavior. This scoping review searched PubMed, Web of Science, IEEE Xplore, and Scopus for English-language studies from 2000 to July 2026 addressing motorsport safety, injury biomechanics, finite element analysis or modeling, protective-system design, or impact-mitigation materials. Study characteristics were charted, and evidence was narratively synthesized following scoping-review principles. Fifty-six studies were included. Injury prevention increasingly draws on head-kinematic field measurements, injury epidemiology, region-specific finite element analyses, and studies of helmets, restraints, seats, and energy-absorbing structures. The most direct evidence concerns integrated head-neck protection, restraint and seat geometry, and the use of finite-element methods to interpret local injury mechanisms. Emerging material strategies are promising but rely mainly on component-level or simulation-based evidence rather than driver-level injury outcomes. Overall, biomechanics, protective-system design, and material selection are increasingly integrated, but evidence remains uneven. Real-world injury surveillance is limited, simulation studies predominate over field outcomes, and direct evidence linking emerging materials to driver-level injury reduction remains sparse. Progress requires better surveillance datasets, stronger motorsport-specific validation, and more driver-centered protective design.

Bioengineering doi: 10.3390/bioengineering13101158

Authors: Hyoyeong Jeong Jinsoo Kim Eui Chul Lee

Respiration is a critical vital sign for assessing the health status of infants during the first year of life. Conventional video-based respiratory monitoring methods often rely on manually defined regions of interest (ROIs) targeting specific anatomical regions, such as the chest or abdomen, or employ pretrained object detection models to localize body regions before extracting respiratory signals. However, in unconstrained sleep environments, infants may adopt diverse sleep postures, making it difficult to assume that respiration-induced movements will consistently appear in the same anatomical region or image location. To address this limitation, we propose an anchor-based respiratory-region localization framework that identifies spatial regions exhibiting prominent respiratory periodicity across the entire video frame without relying on predefined anatomical ROIs or separate body-region localization models. For each grid cell, time-series signals are constructed from inter-frame color/intensity variations and optical-flow-derived motion changes. The magnitude spectra of these signals are then multiplied element-wise to generate a respiratory periodicity map representing the spatial distribution of respiration-related periodicity across the frame. The grid cell with the highest Periodicity Score is selected as the Anchor Grid Cell and used as the primary respiratory region for respiratory-rate estimation. The proposed method was evaluated using 6-fold subject-wise cross-validation on 374 video clips from 15 infants. The single-Anchor Grid Cell approach achieved a mean absolute error (MAE) of 1.9953 breaths per minute (BPM), a root mean square error (RMSE) of 3.9681 BPM, and a Pearson correlation coefficient (PCC) of 0.7287. Moreover, MAEs below 2.5 BPM were obtained for the supine, side-lying, and prone postures, indicating consistent estimation performance across the evaluated sleep postures. These findings suggest that respiration-related information in infants is not necessarily confined to predefined anatomical regions and may instead be expressed at different spatial locations depending on sleep posture.

Bioengineering doi: 10.3390/bioengineering13101157

Authors: Lexi Balster Ramy A. Hinna Ayman A. Hassan Gamal Shiha Ahmed Lotfy

Liver failure and advanced chronic liver disease remain major causes of morbidity and mortality, while donor organ shortages, procedural complexity, cost, and the need for lifelong management after transplantation limit the availability of liver transplantation. Mesenchymal stromal cells (MSCs) have attracted substantial interest as a cell source for liver regeneration because of their accessibility, expansion capacity, immunomodulatory properties, and regenerative secretome. MSC-based hepatic differentiation, however, remains challenging, and cells generated in vitro frequently display an incomplete hepatocyte phenotype and limited maturation. Increasing evidence indicates that soluble differentiation factors alone are insufficient to reproduce the complex microenvironmental cues that regulate hepatic cell fate. Biomaterials can provide three-dimensional architecture, extracellular matrix (ECM) ligands, mechanical cues, cell-adhesive sites, and controlled presentation of growth factors, thereby acting as more than passive carriers of cells. In this mini-review, we summarize studies investigating natural and synthetic biomaterials for MSC hepatic differentiation, with emphasis on collagen-, gelatin-, heparin-, alginate-, and decellularized ECM-based systems as well as synthetic polymeric scaffolds such as poly(lactic-co-glycolic acid) and poly(L-lactic acid)-co-poly(ε-caprolactone). We discuss how biomaterial composition and three-dimensional organization may cooperate with hepatic differentiation media to promote hepatic phenotype and function. We also highlight an important distinction between expression of hepatic markers and acquisition of mature hepatocyte function, and identify opportunities for biomaterial systems that integrate biochemical, biophysical, and paracrine cues to achieve more reproducible hepatic maturation and clinically relevant liver regeneration.

Bioengineering doi: 10.3390/bioengineering13101156

Authors: Takahiro Mizukami Jun Ohashi Hiroki Nishimura Ryota Yokoiwa Yuji Takayanagi Hirotsugu Soga Shintato Nakayama Eisuke Shimizu

Background: Narrowing of the trabecular–iris angle (TIA) is a key anatomical feature of primary angle-closure disease and contributes to impaired aqueous humor outflow; however, conventional angle assessment requires specialized expertise or costly imaging equipment. This study aimed to develop an artificial intelligence model for quantitative TIA estimation from anterior-segment images acquired with a smartphone-attached slit-lamp microscope. Methods: In this retrospective study, 127,512 candidate frames were extracted from anterior-segment videos obtained from 2654 eyes. After automated diagnosability filtering, 59,412 frames suitable for TIA500 estimation were retained and assigned to the training, validation, and test datasets using patient-level grouping. A ResNet50-based model was trained to estimate inferior, nasal, and temporal TIA500 values using the corresponding measurements obtained by anterior-segment optical coherence tomography (AS-OCT) as the ground-truth reference. Model performance was evaluated by comparing the AI-predicted values with the AS-OCT measurements using mean absolute error (MAE), root mean squared error (RMSE), Spearman’s rank correlation coefficient, and the intraclass correlation coefficient (ICC). Results: Per-eye MAEs were 7.01° ± 5.69°, 7.28° ± 5.80°, and 7.90° ± 6.37° for inferior, nasal, and temporal TIA, respectively. Corresponding correlations with reference measurements were 0.683, 0.696, and 0.646, while ICCs were 0.769, 0.764, and 0.743 (all p < 0.001). Conclusions: These findings demonstrate the feasibility of estimating TIA500 from smartphone-based slit-lamp images. However, the moderate correlations with AS-OCT measurements indicate that the present model is not yet sufficiently accurate for independent clinical diagnosis or screening, and further model improvement and external validation are required.

Bioengineering doi: 10.3390/bioengineering13101155

Authors: Yalan Wang Wei Lou Lu Liu Xinmin Lu Qiang Wu Da Long Jianyan Hu

Introduction: This study evaluated whether the refractive prediction accuracy of conventional and new-generation intraocular lens (IOL) power calculation formulas differs between two single-piece acrylic IOL models in eyes with a shallow anterior chamber depth (ACD). Methods: In total, 100 eyes implanted with Softec HD IOLs (Lenstec) and 102 eyes implanted with SN60WF IOLs (Alcon) were included. All eyes had an ACD of ≤3.0 mm. The prediction error (PE) and mean absolute error (MAE) of 11 formulas, with and without optimized constants, were evaluated, including seven new-generation formulas (Barrett Universal II [BUII], Cooke K6 [K6], EVO 2.0, Hill-RBF 3.0, Hoffer QST, Kane, and Pearl-DGS) and four conventional formulas (Haigis, Hoffer Q, Holladay 1, and SRK/T). Results: The new-generation formulas (except K6) showed myopic PEs ranging from −0.108 to −0.181 D in the Softec group; most of these formulas (except Pearl-DGS and Hoffer QST) showed hyperopic PEs ranging from 0.120 to 0.310 D in the SN60WF group. Both before and after optimization, Hill-RBF 3.0 and Pearl-DGS yielded the lowest MAEs in the Softec and SN60WF groups, respectively. Before optimization, most formulas had MAEs > 0.43 D in the Softec group and <0.38 D in the SN60WF group; after optimization, the MAEs were 0.408–0.479 D and 0.334–0.377 D, respectively. Conclusions: The new-generation formulas with ULIB constants showed myopic and hyperopic PEs in the Softec and SN60WF groups, respectively. The SN60WF IOL model demonstrated better refractive prediction accuracy than the Softec model. Formula selection according to IOL model may help improve refractive outcomes in eyes with shallow ACDs.

Bioengineering doi: 10.3390/bioengineering13101154

Authors: Mark M. Vermeulen Pieter F. J. van Altena Dennis Scheidt Angelo Accardo Miriam Menzel

Computational Scattered Light Imaging (ComSLI) exploits the anisotropic scattering on aligned structures to reconstruct biological fiber networks like nerve, muscle or collagen fibers. ComSLI discerns individual fiber pathways also in densely interwoven regions with multiple crossing fibers, achieves micrometer resolution across centimeter fields of view, is compatible with various sample preparations and staining, and only requires an LED light source and a camera, making it attractive for histological tissue analysis. However, while ComSLI precisely determines the 2D (in-plane) fiber orientations, determining the 3D (out-of-plane) fiber inclinations remained challenging due to missing a-priori knowledge of the inclination angles. In this study, we measured differently inclined 3D-nanoprinted fiber bundles with ComSLI, compared the resulting scattering patterns to those obtained from nerve fibers in a tilted brain section (vervet corpus callosum), and analyzed scattering patterns and angular profiles in differently inclined brain regions. We show that the highest scattering intensity occurs along the same curved line as conical diffraction at an inclined grating. With this analytical model, we were able to estimate inclination angles of non-crossing fibers up to 60° inclination with an accuracy of 3° or better. This finally allows to quantitatively determine fiber inclinations and reconstruct 3D fiber pathways with ComSLI.

Bioengineering doi: 10.3390/bioengineering13101153

Authors: Liyun Huang Yuting Tang Xiaobing Chen Zhen Wu Junlan Xiao Qi Xiang Huajian Tian

Since the widespread use of mRNA-based COVID-19 vaccines, significant progress has also been made in mRNA cancer vaccines. Currently, mRNA cancer vaccines represent one of the most promising treatments for advanced tumors, with several candidates in clinical trials. Compared with traditional treatment methods, mRNA cancer vaccines have several advantages. However, several challenges of mRNA cancer vaccines remain, including the accurate and rapid identification of neoantigens and hotspot mutations that match different HLA types. To address these challenges, we review four key aspects of mRNA cancer vaccine development. These include neoantigen identification technologies, machine learning and AI prediction, mRNA sequence and structure design, and the development of LNP delivery systems. Additionally, we analyze the pipelines of leading companies in this field, providing insights into personalized mRNA cancer vaccines. This review also discusses future challenges and prospects, aiming to help guide this promising field toward the full realization of its potential.

Bioengineering doi: 10.3390/bioengineering13101152

Authors: Antony Garcia Adrián Noriega de la Colina Gabrielle Britton Xinming Huang

Black-box models limit the adoption of artificial intelligence in medicine because their predictions are difficult to interpret and reproduce. We present a statistically grounded framework for interpretable, rule-based clinical classification using the Bernoulli Naïve Bayes (BNB) model. Supervised chi-square-guided binarization converts continuous variables into binary indicators by selecting thresholds that maximize association with the clinical outcome within the training folds, which allows BNB to operate on continuous medical data without sacrificing transparency. On three benchmark datasets, Pima Indians Diabetes, Wisconsin Breast Cancer, and Heart Failure Prediction, the framework reached areas under the receiver operating characteristic curve of 0.794, 0.984, and 0.918, respectively. Probabilistic reliability was assessed with a leakage-safe cross-validated calibration analysis reporting Brier score and calibration intercept and slope, and post hoc beta calibration improved probability calibration across datasets. These results indicate that an interpretable, statistically motivated framework can perform comparably to more complex models while providing explicit decision rules expressed in clinical units and risk estimates that are reliable after calibration. A complete worked example further shows that model inference can be reproduced from a printed reference table using only basic arithmetic, without software or proprietary tools, a property that may support auditable use of artificial intelligence in clinical settings.

Bioengineering doi: 10.3390/bioengineering13101151

Authors: Paolo Belardinelli Elena Mongiardini Sonia Turrini Alessio Avenanti

Brain stimulation is progressively evolving from anatomically targeted interventions toward individualized, adaptive, and network-oriented approaches. Recent developments in brain-state-dependent stimulation have demonstrated that neuromodulation can be optimized by synchronizing stimulation with ongoing neural activity, while adaptive cortico-cortical paired associative stimulation (ccPAS) has extended this principle to distributed cortical networks. However, current approaches continue to define brain state primarily through local electrophysiological features, such as oscillatory phase or power, even though ccPAS is specifically designed to modify communication between interconnected cortical regions. In this opinion paper, we argue that the endogenous variable guiding ccPAS should instead describe the instantaneous physiological state of the targeted cortico-cortical pathway. We propose that online source-space functional connectivity provides a biologically meaningful and technically feasible estimate of this pathway state, thereby enabling a new framework of network-state-dependent neuromodulation. As illustrative applications, we consider pathways spanning different levels of experimental characterization, from established sensorimotor interactions to the less-characterized ventral premotor cortex (PMv)-posterior superior temporal sulcus (pSTS) pathway within the action-observation network (AON). This distinction highlights that the proposed framework is pathway-general, while its experimental implementation requires pathway-specific characterization of stimulation parameters and appropriate physiological readouts. In this framework, resting-state EEG is rapidly reconstructed in source space to estimate pathway connectivity, which is subsequently used to personalize stimulation targets and guide state-dependent ccPAS of a functionally defined cortico-cortical circuit. Finally, we discuss the technological prerequisites, methodological challenges, biological rationale, and open scientific questions that must be addressed before pathway-state-guided neuromodulation can be experimentally validated. We argue that redefining brain state as the instantaneous physiological state of the communication pathway targeted by stimulation may represent the next conceptual step in brain stimulation.

Bioengineering doi: 10.3390/bioengineering13101150

Authors: Augusto Ielo Lilla Bonanno Angelo Quartarone Serena Dattola

Freezing of gait (FoG) in Parkinson’s disease is clinically heterogeneous, yet wearable-sensor studies predominantly focus on binary FoG detection. We investigated whether right-ankle inertial signals can characterize clinically meaningful FoG manifestations within a three-stage conditional framework. Accelerometer and gyroscope data from 22 participants in the FoG-STAR dataset were segmented into 1 s windows. Three subject-independent classification stages were evaluated separately: locomotor non-FoG versus FoG, kinetic versus akinetic FoG, and shuffling versus trembling. Nine interpretable biomechanical features were compared with Logistic Regression, RBF-SVM, Random Forest, and a raw temporal 1D-CNN using frozen participant-independent partitions. The CNN yielded a mean Stage-1 balanced accuracy (0.877 ± 0.034), whereas for Stage-2 kinetic–akinetic discrimination, Logistic Regression and the CNN yielded mean balanced accuracies of 0.730 ± 0.044 and 0.727 ± 0.065, respectively. Stage-3 performance was lower and more variable (Logistic Regression, 0.666 ± 0.133). Nested leave-one-subject-out analysis restricted to mixed-phenotype participants further highlighted subject-level heterogeneity. Kinetic FoG showed greater residual movement amplitude and movement-rate characteristics than akinesia. Protocol-observed locomotor FoG burden was positively associated with FoG-Q (ρ=0.536) and MDS-UPDRS III (ρ=0.785). Wearable inertial signals therefore showed clinically interpretable differences among FoG manifestations, while finer phenotype discrimination remained heterogeneous and requires external validation in independent cohorts.

Bioengineering doi: 10.3390/bioengineering13101149

Authors: Adrian Barbilian Marius Moga Mark Edward Pogarasteanu Vlad Tau Alex Lucaci Constantin-Andreas Gorgan Alexandru Diaconu Robert Daniel Dobrotă Razvan Adam Cristian Stoica

Background and objectives: Patient-specific skeletal implants are designed through a pipeline of imaging, computational planning and, increasingly, artificial intelligence. Each pillar has its own validation literature, but no review has asked what evidence supports each one in implant design itself. We mapped that evidence from the biomedical-indexed literature in this paper. Methods: A systematic review was conducted under a protocol committed before retrieval (PRISMA 2020, PRISMA-S, SWiM). MEDLINE, Europe PMC, Crossref and OpenAlex were searched on 2 and 3 September 2026 for primary studies from 2015 onwards applying artificial intelligence, an evaluated imaging pipeline or computational planning to an implant whose geometry derives from the individual patient’s anatomy. Engineering indexes were unavailable, so this review maps biomedical-indexed literature, and a bounded probe was used to estimate what that omits. Two artificial intelligence screeners from different vendors screened every record; every clinical study was appraised twice, independently, by assessors from different vendors. Claims about validation quality are reported on the studies read in full. All screening, charting and appraisals were machine-performed and no human-verified subsample exists. Results: Of 5059 records, 639 studies were included: 373 technical studies, 179 clinical series, 75 comparative studies and 12 randomised trials. Computational planning was applied in 601, an evaluated imaging step in 219 and artificial intelligence in 36. Randomised evidence came from 12 small trials, 10 of them craniomaxillofacial, which evaluated the delivered implant rather than the planning method; virtual surgical planning was used in 11 of them. When restricting to the four studies that measure deviation from the virtual plan, with smaller meaning better, patient-specific implants deviated less from the plan (standardised mean difference −1.21, 95% CI −1.69 to −0.74). Revision or removal was less frequent (3 of 135 against 25 of 149; risk ratio 0.28, 95% CI 0.11 to 0.72); operative time, complications and infection gave intervals compatible with a difference either way. No randomised trial was at low risk of bias, and 69 of 75 comparative studies were at serious risk, mostly of unadjusted confounding, so all estimates are of very low certainty. Among technical studies read in full, independent validation material was used in fewer than half, and files were shared in about one in fourteen. Of 36 artificial intelligence studies, 4 were evaluated on independent data, 2 were used for a real patient, none were clinically comparative, and 8 of 17 cranial studies relied on AutoImplant family benchmarks. Conclusions: The design pipeline is unequally validated: planning has clinical evidence concentrated in one region, computational methods are widely applied but rarely validated independently or shared, and artificial intelligence has benchmark evidence for cranial defect completion and little else. Outcome studies of patient-specific implants should be read as evidence about whole pipelines, and future studies should evaluate the components.

Bioengineering doi: 10.3390/bioengineering13101148

Authors: Khulood Abu Maria Sara Abu Tarboosh Adi El-Dalahmeh Ola Tarawneh

Three-dimensional (3D) bioprinting of hydrogel inks is an enabling technology for tissue engineering and the fabrication of personalized drug-delivery systems. However, in most laboratories the extrusion parameters are still optimized by slow, resource-intensive trial-and-error, because printability depends on the combined and often nonlinear influence of ink composition, extrusion pressure, and nozzle geometry rather than on any single variable. Alginate–methylcellulose (Alg–MC) is a widely used pharmaceutical bioink whose shear-thinning behavior makes this optimization particularly demanding. Here we developed a physics-informed machine-learning framework that predicts extrusion print quality before printing. A hybrid dataset of 3280 labeled samples was assembled by combining laboratory extrusion experiments, computational fluid dynamics (CFD) simulations of nozzle flow, and controlled augmentation with a deep belief network (DBN). Five physics-informed features encoding pressure-to-diameter and shear-rate interactions were engineered to embed rheological domain knowledge. Five base classifiers alongside two ensemble strategies, seven models in all, were benchmarked under five-fold stratified cross-validation. The voting ensemble performed best (accuracy, 93.3%; F1 = 0.929), narrowly ahead of the stacking ensemble (93.3% accuracy; F1 = 0.928). Alg:MC ratio, extrusion pressure, and the pressure–diameter interaction term stood out as the dominant predictors, consistent with established alginate rheology. On external validation—withholding one composition ratio and one needle gauge entirely from training and testing on those genuinely unseen conditions—accuracy was 75.9%, a stricter test than a random held-out split. Overall, the framework offers a practical, interpretable way to cut down experimental burden in pharmaceutical bioink development.

Bioengineering doi: 10.3390/bioengineering13101147

Authors: Xudong Wang Qinghong Gao Ran Yang Xuelian Yang Rui Yao Shu’an Liu Maosong Jiang Caiye Fan Zuoping Tan Yuanyuan Wang

Optical Coherence Tomography Angiography (OCTA) provides clear visualization of ocular microvascular details, serving as a critical tool for assessing retinal and choroidal vascular systems. However, the complex capillary structure and low capillary-to-background contrast in high-resolution OCTA images pose significant challenges for segmentation. To address the low segmentation accuracy of traditional methods, this paper proposes RCNet, a hybrid CNN-Transformer network. RCNet adopts an encoder–decoder architecture based on Bi-Level Routing Attention (BRA), where bi-level routing attention focuses on densely vascularized regions, optimizing global capillary feature extraction. A Convolutional Bottleneck Module (CBM) splits bottleneck features into two channel groups using a fixed ratio α. A high-capacity branch enhances local details, while a lightweight branch preserves complementary information. The branches are then fused. Additionally, a Dense Skip Connection Module (DSCM) enhances feature information flow in low-contrast regions, compensating for spatial information loss. By integrating CNN’s local feature extraction with Transformer’s global dependency modeling, RCNet achieves accurate segmentation of OCTA capillary images. Experiments on the OCTA-500 and ROSE datasets demonstrate that RCNet achieved a higher point estimate under the current protocol, facilitating and assisting in the clinical assessment of ocular diseases.

Bioengineering doi: 10.3390/bioengineering13101146

Authors: Zhaoyan Zhu Bo Yin Pengyu Jiang Yue-Ting Sun

Automatic sleep staging relies on effective representation of physiological signals. Representative raw-signal sleep-staging encoders typically transform input waveforms while progressively compressing their temporal resolution within the same encoding path. However, explicit modality-specific, sample-aligned, and task-driven signal-domain transformation before epoch-level feature extraction remains comparatively underexplored. To investigate this design, we propose RefineSleepNet, an end-to-end sequence-to-sequence framework that explicitly separates these two stages. Separately parameterized Signal Refinement U-Net (SRU-Net) branches first transform EEG and EOG waveforms while preserving the original temporal length, after which Epoch Feature Extractors (EFEs) extract epoch-level representations from the refined signals. Fused Sequence Attention (FSA) then performs multimodal integration and cross-epoch contextual modeling before epoch-wise classification. All stages are jointly optimized using the final sleep-staging objective without a separate waveform reconstruction target. RefineSleepNet was evaluated on Sleep-EDF-39, Sleep-EDF-153, and ISRUC-3 using subject-independent 10-fold cross-validation, achieving accuracy/macro-F1 (MF1) scores of 84.5%/79.8%, 83.5%/78.1%, and 80.3%/78.1%, respectively. Ablation experiments supported the coordinated design, while statistical analysis and intermediate-representation evaluations further examined the robustness and characteristics of the learned transformation. The learned transformation retained broad temporal and spectral organization while introducing local waveform changes and modality-specific spectral redistribution. These results support the staged organization as a viable modeling strategy for EEG/EOG sleep staging.

Bioengineering doi: 10.3390/bioengineering13101145

Authors: Sang-Hyeon Gill Hyun-Young Choi Yuseop Kim

CT pulmonary angiography (CTPA) is the reference test for pulmonary embolism (PE), whereas AI datasets often contain display-domain (non-DICOM) non-contrast CT (NCT). We present a three-stage pipeline: a representation-consistency gate calibrated on DICOM-verified images (Stage 1), unpaired CycleGAN translation (Stage 2), and a frequency-split composition keeping the acquired finest spatial band plus support-guided residual refinement (Stage 3). Of 208 patients, 30 were held out. The final output reduced mediastinal distribution distance (14.92 to 7.19) and contrast-gap error (11.00 to 4.25; Holm-adjusted p < 0.001). Distribution metrics did not differ significantly from CUT, whereas three edge-based metrics favored the proposed method in all 30 patients. On 24 external NCT scans from three manufacturers (full-pipeline feasibility analysis without paired CTPA), distribution distances decreased and edge metrics again favored the method. The gate flagged all degraded external CTPA images (blur, down-sampling, heavy compression) in every calibration resample. An external 3D CTPA-trained PE model labeled unopacified arteries as embolus on NCT; enhancement reduced this volume in normal patients (12.0 to 1.7 mL), but detection stayed at chance, as on the original NCT (AUROC 0.47; original 0.42; CTPA 0.85). The pipeline harmonizes display-domain NCT with limited structural alteration for dataset preparation rather than CTPA-based detection.

Bioengineering doi: 10.3390/bioengineering13101143

Authors: Maksim Solopov Roman Ishchenko Viktor Turchin Viktoria Bushe Anna Kavelina Gulnara Akopian Elizaveta Lebed Andrey Popandopulo

Combined iron oxide magnetic nanoparticles (MNPs) and external magnetic fields (MFs) are explored to direct osteogenic differentiation of mesenchymal stem cells (MSCs) in bone tissue engineering, yet exposure parameters vary widely. This systematic review and meta-analysis searched PubMed and Dimensions, with Consensus and SciSpace as exploratory scoping aids (final search 22 November 2025). Twenty-five studies entered qualitative synthesis and twenty-two quantitative synthesis (‘MNP + MF’ vs. ‘MNP without MF’; standardized mean difference [SMD] as Hedges’ g; restricted maximum likelihood random-effects model with Hartung–Knapp–Sidik–Jonkman confidence intervals [CIs]). Across nine outcomes, effect directions were predominantly positive; uncorrected significance held only for alkaline phosphatase (ALP) activity (SMD = +4.42; 95% CI 0.97–7.86; k = 14) and mineralization (SMD = +2.41; 0.03–4.80; k = 7). After Holm correction, no outcome remained significant. Heterogeneity was critical (I2 > 75% for five outcomes). Funnel–plot asymmetry for ALP suggested possible small-study effects; trim-and-fill reduced the pooled SMD to +1.34 (−1.60–4.27). Findings constitute a low-certainty, hypothesis-generating signal rather than confirmed efficacy; standardized protocols and blinded designs are needed before translational claims.

Bioengineering doi: 10.3390/bioengineering13101144

Authors: Na Li Yuanbo Gao Liang Lu Yu Liu Zhenbin Zhang Xing Zhang

Although calcium alginate microspheres prepared by microfluidic methods have been widely investigated, the combined effects of dispersed-phase composition, continuous-phase properties, and surfactant combinations on microsphere formation remain insufficiently understood. In this study, a microfluidic system was established using sodium alginate as the aqueous phase, mineral oil as the continuous phase, and calcium chloride as the receiving phase. Response surface methodology was applied to model and optimize the key parameters involved in microsphere formation. The regression model showed good fit and reliable predictive ability. Under the optimized conditions, validation experiments yielded an average target-size fraction of 81.25%, which agreed well with the predicted value, demonstrating the feasibility and reproducibility of the optimized process. Biological evaluation showed that the optimized alginate microspheres exhibited no significant cytotoxicity toward BxPC-3 cells and maintained high cell viability, as confirmed by Live/Dead staining and CCK-8 assay. Confocal imaging further showed that the encapsulated BxPC-3 cells remained highly viable and formed three-dimensional clusters within the alginate matrix. These findings provide a reproducible microfluidic approach for generating calcium alginate microspheres with potential utility in cell encapsulation, three-dimensional culture, and related biomedical applications.

Bioengineering doi: 10.3390/bioengineering13101141

Authors: Jiaming Shi Yongze Jiang Wanqi Xiong Tao Zhang Yujin Dong Xinli Dong Jiahe Liu Baoyi Liu

This study investigated the biomechanical effects of distal fragment translation guided by target postoperative first–second intermetatarsal angle (IMA) in minimally invasive transverse osteotomy (MITO) for hallux valgus using patient-specific finite element models. Three-dimensional models were reconstructed from computed tomography data of three female patients representing mild, moderate, and severe hallux valgus. The distal first-metatarsal fragment was translated laterally to achieve target postoperative IMA values of 9°, 8°, 7°, 6°, and 5°. Maximum von Mises stresses of the first through fifth metatarsals and Kirschner wires, together with mean osteotomy contact pressure and contact area, were evaluated under standardized static standing. As the target IMA decreased, the required translation increased, osteotomy contact area decreased, and mean contact pressure increased in all models. First-metatarsal stress increased most markedly in the severe model, reaching 25.74 MPa at 5°. In the mild and moderate models, further correction at the smaller IMA settings produced relatively limited additional changes in metatarsal stress while fixation stress and osteotomy-interface demands continued to increase. In contrast, the severe model showed continued increases in first-metatarsal and fixation stresses with further correction. These exploratory findings indicate that the mechanical effects of correction depend on both the target postoperative IMA and the distal fragment translation required to achieve it.

Bioengineering doi: 10.3390/bioengineering13101142

Authors: Longxiang Li Longzhou Hua Mingwei Liu Ye Luo Yingjie Lyu Fei Tian Shaobai Wang

Background: Knee injuries are common during military training, and fatigue alters lower-limb movement patterns, increasing injury risk. However, little is known about fatigue effects on knee kinematics during military marching. Conventional gait analysis is limited to sagittal plane motion, leaving multi-planar knee function poorly understood. Objective: This study investigated six-degree-of-freedom (6DOF) knee kinematic changes in healthy male military personnel marching at 8 km/h on a treadmill before and after fatigue, focusing on initial contact (IC), stance-phase flexion peak (SPF), and stance-phase extension peak (SPE). Methods: Thirty-eight healthy male soldiers marched on a treadmill at 8 km/h. A motion capture system (infrared stereo cameras + reflective markers) collected 6DOF kinematic data. Fatigue was induced by loaded marching (15 kg vest), with the endpoint defined as Borg RPE ≥ 18 or heart rate ≥ 90% of age-predicted maximum. Post-fatigue data were collected at the same speed. Flexion, rotation, and anterior–posterior displacement at IC, SPF, and SPE were compared pre- vs. post-fatigue. The Holm–Bonferroni correction was applied for multiple comparisons. Results: After Holm–Bonferroni correction, knee flexion angles decreased significantly at SPF (3.64°, p = 0.009, Cohen’s d = 0.52) and SPE (2.75°, p = 0.032, d = 0.41), with a non-significant trend at IC (2.53°, p = 0.086). Internal rotation shifted toward external rotation at IC (2.48°, p = 0.039, d = 0.46) and SPE (2.99°, p = 0.036, d = 0.54), with a non-significant trend at SPF (p = 0.076). No other kinematic changes, including anteroposterior displacement, exceeded the system’s measurement error threshold or remained significant after correction. Conclusions: Fatigue reduces knee flexion during the stance phase at 8 km/h (with the most robust changes at SPF and SPE) and shifts internal–external rotation toward external rotation at IC and SPE. These 6DOF kinematic changes may compromise knee stability and increase ACL and patellofemoral injury risk.

Bioengineering doi: 10.3390/bioengineering13101140

Authors: Xiaoning Bi Hanna Gootin Wenyue Su Michel Baudry

Extracellular vesicles (EVs) are increasingly recognized as regulated mediators of intercellular communication in the central nervous system (CNS), transferring proteins, lipids, metabolites, and nucleic acids among neurons, glia, and other cells. EV communication comprises a complex integrated pathway that includes vesicle biogenesis, selective cargo sorting, secretion, extracellular transport, target-cell recognition, uptake, intracellular trafficking, functional cargo delivery, and recipient-cell responses. These processes are controlled by membrane trafficking machinery, cytoskeletal dynamics, lipid metabolism, calcium signaling, and the endosomal–lysosomal system and are further shaped by the physiological and pathophysiological states of donor and recipient cells. Genetic variation, proteotoxic stress, inflammation, metabolic and redox alterations, neuronal activity, and endolysosomal dysfunction can therefore affect multiple stages of the EV communication pathway in neurodevelopmental and neurodegenerative disorders. EV alterations reflect dynamic states rather than fixed disease signatures, with effects depending on cellular origin, EV subtype, anatomical context, developmental or disease state, and recipient-cell competence. This review examines CNS EV physiology and proposes a state-dependent framework for biomarkers and therapies, distinguishing EV detection from functional cargo delivery, cross-sectional associations from longitudinal progression, and cellular uptake from intracellular target engagement.

Bioengineering doi: 10.3390/bioengineering13101139

Authors: Ciro Gargiulo Isacco Van Hung Pham Kieu C. D. Nguyen Toai Cong Tran Emilio Jirillo Luigi Santacroce

Background: Carcinogenesis has traditionally been attributed to the accumulation of somatic mutations in nuclear genes regulating cell proliferation, survival, and apoptosis. Although hereditary genetic predispositions account for less than 10% of cases, the extensive heterogeneity of the somatic mechanisms driving tumorigenesis—including subsets of neoplasms lacking canonical nuclear mutations—suggests that non-mutational mechanisms and organelle dynamics significantly influence tumor onset. Emerging evidence increasingly supports the concept that mitochondrial dysfunction and metabolic reprogramming are essential factors in tumor development, rather than merely downstream consequences of nuclear genetic alterations. While current literature establishes mitochondrial alterations as crucial co-initiators and facilitators of early transformation, defining whether metabolic failures may eventually act as primary trigger remains an ongoing area of investigation. Within this perspective, the “Endosymbiotic/Evolutionary Theory” provides a useful heuristic model, since it conceptualizes metabolic rewiring not as a literal genetic reversal, but as an adaptive shift toward ancestral, highly glycolytic survival programs when oxidative microenvironments are compromised. Methods: We conducted a comprehensive theoretical and comparative analysis of mitochondrial biology, interrogating the structural, functional, and evolutionary parallels between mitochondria, intracellular bacteria, and viruses within the conceptual context of the endosymbiotic theory. Particular attention was given to mitochondrial proteins associated with organellogenesis, bioenergetic control, and molecular mechanisms facilitating mitochondrial–nuclear communication, including the integration of mitochondrial DNA (mtDNA) into nuclear DNA (nDNA) as nuclear mitochondrial DNA segments (NUMTs). Results: Under sustained bioenergetic stress, mitochondrial retrograde signaling can achieve regulatory dominance over the nucleus, driving a functional dissociation between organellar metabolic state and nuclear control. While not autonomous in a strict sense, this shift echoes the evolutionary legacy described by the endosymbiotic theory, whereby mitochondria act as semi-independent bioenergetic systems capable of imposing profound effects on nuclear gene expression and cellular fate. The resulting state is characterized by metabolic reconfiguration, loss of differentiation fidelity, and the acquisition of cancer stem cell–like properties. This process may unfold through a sequence of interrelated events: (1) persistent endogenous or exogenous stressors that challenge cellular homeostasis; (2) establishment of a chronic inflammatory state and sustained redox imbalance; (3) progressive disruption of the cellular microenvironment and inter-organelle communication; (4) emergence of a permissive pre-neoplastic niche characterized by impaired oxidative phosphorylation (OXPHOS), altered mitochondrial dynamics, and sustained retrograde signaling; (5) activation of adaptive survival pathways and escape from canonical cell death and checkpoint controls; and (6) progression toward a dedifferentiated, stem-like cellular phenotype associated with metabolic plasticity and increased resilience. Conclusions: We recommend that cancer be reframed as a dedifferentiated cellular state primarily driven by mitochondria-derived signaling, reflecting the organelle’s ancestral autonomy and the reactivation of evolutionarily conserved mechanisms involving transposons, retrotransposons, Long Interspersed Nuclear Element-1 (L1), transposable elements (TEs), reactive oxygen species (ROS), and mtDNA–nDNA crosstalk, including the generation of nuclear mitochondrial DNA segments (NUMTs). This perspective extends the canonical view of predominantly nuclear-driven oncogenesis by reframing carcinogenesis as a bioenergetic and bio-evolutionary process, thereby highlighting mtDNA–nDNA interactions and mitochondrial signaling as central targets for cancer prevention, diagnosis, and therapy.

Bioengineering doi: 10.3390/bioengineering13101138

Authors: Michael Baldock Silvia Caggiari Antonio Capozio David E. Lunn Frances Gawne Ioannis Delis Sarah L. Astill

Background: End-effector robotic devices are increasingly used in rehabilitation to retrain neural pathways through repetitive, task-specific exercise. However, the influence of movement kinematic variables remains unclear. This systematic review aimed to; (i) identify end-effector robotic systems used in upper- and lower-limb spinal cord injury rehabilitation, (ii) describe their technical specifications and operational use, and (iii) explore associations between device use characteristics and study outcomes. Methods: A systematic review of five databases was conducted from inception to June 2026. Results: Fifteen studies, ten lower-limb and five upper-limb, met the inclusion criteria. These covered five commercially available lower-limb devices and three upper-limb devices. Most lower-limb studies demonstrated moderate-to-large improvements in balance, ambulation (WISCI-II), walking endurance (6mWT), and independence (SCIM-III). End-effector training produced comparable outcomes to both conventional therapy and exoskeleton-based rehabilitation in controlled comparisons. Upper-limb studies showed substantial heterogeneity in outcome selection, reflecting diverse rehabilitation goals. Conclusions: WISCI-II and 10MWT were the most used lower-limb functional assessments, with SCIM-III used to assess independence for both upper- and lower-limb motor impairments. Reporting of kinematic settings are often limited and concentrated at study onset, limiting assessment of their contribution to outcomes. Reporting the progression of session-level metrics is needed to better understand how time-varying kinematics affect intervention effectiveness.

Bioengineering doi: 10.3390/bioengineering13101137

Authors: Ahmed Fekry Salman Jamal I. Al-Nabulsi Ali Olamat Faten H. Abdelaziem Hoda Abd E. A. El-Talawy Ahmed E. Fayed

(1) Background: Balance determines functional status in cerebral palsy (CP), yet its predictability from clinical impairment measures remains unclear. This study aims to evaluate the predictability of three balance targets using a clinical impairment battery and determine whether standing balance comprises discrete and graded components. (2) Methods: A total of 173 children with CP (aged 0.9–9.0 years) were assessed using the Early Clinical Assessment of Balance (ECAB) (ECAB total, Part 1: head/trunk control, Part 2: standing). Spasticity, muscle strength, posture, tone asymmetry, age, sex, and CP subtype were analyzed as predictors using cross-validated regression models and a two-part hurdle model. (3) Results: Head and trunk control demonstrated high predictability (R2 = 0.807), while the ECAB total score achieved R2 = 0.671. Standing balance showed moderate predictability in standard regression (R2 = 0.519). The hurdle model decomposed this: it discriminated standing capability with high accuracy (AUC = 0.982) but explained only 41.4% of graded variance among standers (R2 = 0.414), clarifying rather than improving overall prediction. Muscle strength was the primary predictor across all models. (4) Conclusions: Within this cohort, clinical impairments reconstruct head and trunk control and predict whether a child can stand but do not explain standing quality; external validation is still required. The ECAB standing subscale captures graded information not reconstructable from the impairment battery as specified (which by design excludes global severity classification) and should be measured directly.

Bioengineering doi: 10.3390/bioengineering13101136

Authors: Som V. Thomas Jing Chen Dong Qian Paul Spearman Donglu Shi

SARS-CoV-2 has adapted continuously since 2019, yet no multi-parameter framework has characterized this trajectory across geographically distinct United States populations. We present a systematic 13-parameter analysis of 137,059 high-quality genomes from ten diverse US states across three phases: pre-vaccination (January–December 2020), early vaccination (January–June 2021), and sustained immune pressure (July 2021–December 2025), the latter dominated by the high-volume Delta period with an early-Omicron contribution. Approximately 5000 genomes were retrieved per state per phase from GISAID EpiCoV under stringent quality filters, and each phase dataset was partitioned into three subsets, giving 90 subsamples in total, to quantify subsampling stability. Parameters spanned genome-wide mutation rate, dN/dS and Ti/Tv, site-specific selection (HyPhy FUBAR), spike and ACE2 contact-residue profiling, insertion–deletion characterization, furin cleavage site evolution, N-linked glycosylation dynamics, lineage–mutation concordance, amino acid physicochemical shifts, and protein-wise dN/dS ranking and mutation load. Genome-wide mutation rate rose monotonically in all ten states, from 0.050% to 0.149%, and dN/dS rose from 0.611 to 1.103, crossing neutrality in nine of ten states by Phase 3. Amino acid mutation load increased 3.95-fold, from 9.12 to 36.03 substitutions per sequence, while Ti/Tv fell from 3.69 to 2.11. Spike position 142 was diversifying in all ten states in Phase 3, with positions 484, 452, and 478 recurring in 70, 60, and 60 percent of states. ORF8 followed a spike-then-collapse trajectory (dN/dS 3.43, 14.23, 1.01), consistent with progressive loss of function. The reproducibility of these effects across ten geographically distinct state populations establishes a state-level baseline for SARS-CoV-2 immune escape.

Bioengineering doi: 10.3390/bioengineering13101135

Authors: Ricardo Batista André Moreira Rui B. Ruben Maria Helena Figueiral Nuno Alves João Carlos Sampaio-Fernandes

Preload loss and plastic deformation of titanium abutment screws under repeated tightening are clinically significant yet incompletely characterized across systems with differing head geometries and surface treatments. Two ISO 14801-certified titanium screws were evaluated: hexalobular-head gold-plated (Gold-Tite®; 3i Zimmer Biomet, now ZIMVIE) and anodized star-shaped (Vega®; Klockner®). Six groups (n = 5 screws) varied by screw type, torque (20–40 N·cm), and cycle count (10 or 20). Reverse torque value (RTV) was recorded at every cycle and analyzed at the screw level; a linear mixed-effects model (screw as random effect) characterized how RTV evolved across cycles, with screw-level non-parametric tests for between-protocol contrasts. Micro-CT (19.9 μm; n = 3 per group) used Otsu segmentation. RTV was lower than insertion torque in every screw (30/30; 73–89% of applied torque). RTV declined significantly across cycles in the Vega/Klockner system (−0.17 N·cm per cycle, p = 0.006) but not in the ZFX/ZIMVIE system (p = 0.302), which showed greater between screw variability. Between protocol contrasts at n = 5 were null or underpowered. Micro-CT showed head localized change in both systems (mean cross-sectional area essentially unchanged); in the ZFX/ZIMVIE screws, this change appeared specifically under the over-use condition, whereas the Klockner screws showed comparable head change across all protocols. Preload loss is intrinsic to abutment screw tightening irrespective of design; RTV behaviour under repeated cycling is system-dependent, and abutment screws should be treated as limited-use components.

Bioengineering doi: 10.3390/bioengineering13101133

Authors: Stefano Bellucci

Artificial intelligence (AI) is increasingly used in biosensing, yet its role is often limited to post-processing or high-accuracy regression on simulation-generated datasets. This critical narrative review examines physics-informed and explainable AI for nanomaterial-based biosensors, emphasizing graphene and other two-dimensional materials, surface plasmon resonance (SPR), terahertz (THz) metasurfaces, electrochemical platforms, field-effect transistors, surface-enhanced Raman spectroscopy, and wearable systems. Recent peer-reviewed studies and relevant technical guidance were critically compared with respect to sensing physics, data provenance, AI task, interpretability, validation strategy, and evidence level. Across these modalities, AI supports forward surrogate modeling, inverse design, spectral interpretation, classification, calibration, and uncertainty-aware decision support, but predictive accuracy alone does not establish translational maturity. Particular attention is given to explainable AI, the distinction between physics-guided and genuinely physics-informed learning, small-data validation, fabrication tolerance, drift, and the simulation-to-experiment gap. Quantitative comparisons place reported SPR and THz sensitivities in context, while independent experimental studies provide benchmarks for real-device and biomedical validation. Clinically credible intelligent biosensors should combine mechanistic constraints, uncertainty quantification, device- and batch-aware validation, interpretable features, realistic biological matrices, and prospective evaluation.

Bioengineering doi: 10.3390/bioengineering13101134

Authors: Hairul Aysa Abdul Halim Sithiq Liyana Shuib Muneer Ahmad Chermaine Deepa Antony Mohsin Iftikhar Peer Azmat Shah

Lung diseases such as Honeycombing (HC) and Ground Glass Opacity (GGO) present significant challenges in medical image analysis due to their overlapping visual characteristics and complex structural patterns in Computed Tomography (CT) images. These conditions often exhibit similar appearances to normal lung tissue, making accurate differentiation difficult and highly dependent on imaging quality and expert interpretation. This paper presents a systematic literature review of studies published between 2011 and 2026 on lung disease detection and classification using medical imaging. A total of 98 selected studies is analyzed to examine recent developments in Machine Learning (ML) and Deep Learning (DL) techniques for automated lung disease analysis. The review covers key aspects including imaging modalities such as CT and chest X-ray, publicly available and private datasets, preprocessing techniques, feature extraction methods, classification models, and evaluation metrics. The findings indicate that traditional machine learning approaches were widely used in earlier studies, while recent research has increasingly shifted towards deep learning methods due to their ability to automatically learn discriminative features from medical images. Among these approaches, Convolutional Neural Networks (CNNs), ResNet-based architectures, DenseNet, and EfficientNet are the most commonly adopted models for lung disease classification. Despite significant advancements, several challenges remain, including limited availability of publicly accessible datasets, class imbalance, variability in imaging protocols, and lack of standardized evaluation frameworks. In addition, misclassification between visually similar conditions such as Honeycombing and GGO remains a persistent issue that affects diagnostic reliability. The review identifies key research gaps and highlights the need for standardized datasets, improved preprocessing techniques, and consistent evaluation protocols. Future research should focus on enhancing model generalisation and robustness across diverse datasets to support reliable automated lung disease diagnosis in clinical practice.

Bioengineering doi: 10.3390/bioengineering13101132

Authors: Kattia Orozco-Romero Jonas Pariona-Torres Jose Cornejo

Type 2 diabetes mellitus (T2DM) affects hundreds of millions of people worldwide, and nearly half of all cases remain undiagnosed. This technical review, conducted under conducted under the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and structured around the Population, Intervention, Comparison, Outcome and Context (PICOC) framework, identifies and synthesises the evidence on artificial intelligence (AI) algorithms for the early detection of T2DM and the risk stratification of its complications, evaluating their performance against conventional diagnostic methods and characterising their clinical implementation contexts. A systematic search in Scopus with a four-stage screening process, synthesised across five research dimensions, yielded 255 records, of which 47 studies met the inclusion criteria (2025–2026). Ensemble methods based on gradient boosting, particularly particularly extreme gradient boosting (XGBoost), emerged as the dominant primary modelling approach, while logistic regression was the most frequently reported algorithm, predominantly as a baseline comparator, with area under the receiver operating characteristic curve (AUC-ROC) values ranging from 0.63 to 0.97. AI models consistently outperformed glycated haemoglobin (HbA1c), whose reported sensitivity for prediabetes was only 49%, reaching accuracies of up to 96.7%, although that upper figure comes from a narrowly defined cohort with a high baseline prevalence and does not represent general screening performance. Multimodal approaches achieved the highest discrimination, combining wearables, continuous glucose monitoring, and genomics (AUC 0.96 internal, 0.90 external, for the separation of self-reported glycaemic states), and a real-world clinical deployment integrated into electronic health records improved glycaemic control outcomes among high-risk patients. An assessment of the eight anchor studies with thePrediction model Risk Of Bias ASsessment Tool updated for AI (PROBAST + AI) found high concern regarding quality in six of them, driven mainly by cross-sectional designs and self-reported outcomes. Overall, AI stands out as a viable and scalable complement to conventional screening, although gaps remain in multicentre prospective validation, reporting standardisation, and algorithmic equity across populations.

Bioengineering doi: 10.3390/bioengineering13101131

Authors: Felix Förtsch Johannes Ries Tom Kasperlik Manfred Wichmann Ragai Edward Matta

The objective was to evaluate the trueness of interocclusal registration obtained with three intraoral scanning systems versus a conventional impression and bite registration in fully dentate patients. Twenty adults with complete dentitions and Angle Class I occlusion were enrolled. Complete-arch scans were obtained with Primescan, iTero Element, and TRIOS 4, followed by bilateral virtual interocclusal registration. Conventional impressions and polyvinyl siloxane records were used to fabricate stone casts, digitized with a laboratory scanner with the record and after manual articulation in maximum intercuspation. Industrial scanner datasets served as participant-specific references. Three-dimensional deviations were analyzed by repeated-measures ANOVA and Holm-adjusted paired t-tests (α = 0.05). The conventional workflow with an interocclusal record showed the greatest deviation (797 ± 308 μm; p < 0.001). Manual articulation (331 ± 167 μm) did not differ significantly from Primescan (262 ± 120 μm), iTero Element (256 ± 101 μm), or TRIOS 4 (325 ± 168 μm) (p ≥ 0.223); no differences were found among scanners under the conditions of this study (p ≥ 0.468). Virtual interocclusal registration showed greater trueness than the conventional workflow with a physical record, whereas manual articulation performed comparably to digital workflows. These conclusions apply only to fully dentate patients with stable occlusal support and maximum intercuspation.

Bioengineering doi: 10.3390/bioengineering13101130

Authors: Ramadhan Ahmed Kamini Narendra Reddy Mahad Mohamed Nakul S. Shekhawat

Purpose: This study aimed to compare the repeatability and agreement of corneal tomographic measurements between the Anterion Cornea App and Pentacam HR in keratoconus (KCN). Methods: Eyes with KCN were imaged three consecutive times with each device. Intra-device repeatability was assessed using intraclass correlation coefficients (ICCs), within-subject standard deviation (Sw), coefficient of variation (CV), and repeatability coefficients (RCs). Inter-device agreement was evaluated using concordance correlation coefficients (CCCs) and Bland–Altman analysis. Parameters included anterior and posterior keratometry values (K1, K2, Kmean, Kmax, and ΔK), central corneal thickness (CCT), and thinnest pachymetry (TP). Results: Twenty-eight eyes from 17 patients were included. Both devices demonstrated high same-session repeatability (ICC > 0.90 and CV < 5%) for all parameters except anterior ΔK and posterior ΔK. Anterion demonstrated lower measurement variability than Pentacam for all parameters. For anterior Kmax, Anterion had an RC of 1.38 diopters (D) versus 2.48 D for Pentacam. Pachymetric RCs were lower with Anterion than with Pentacam (CCT: 6.24 vs. 18.85 μm; TP: 5.93 vs. 20.34 μm). Inter-device agreement was substantial (CCC > 0.95) for most parameters except ΔK. Compared with Pentacam, Anterion measured CCT as 6.16 μm thinner and TP as 11.99 μm thinner. Bland–Altman analysis showed greater inter-device variability for all anterior keratometry parameters, posterior K1, and posterior Kmean. Conclusions: Anterion demonstrated lower same-session measurement variability than Pentacam HR, particularly for Kmax and pachymetry. Future longitudinal studies across multiple sessions are needed to assess whether Anterion’s lower same-session variability confers any advantages in KCN detection and progression monitoring. Overall, the two devices should not be used interchangeably for KCN assessment.

Bioengineering doi: 10.3390/bioengineering13101129

Authors: Stefano Bellucci

Wearable and implantable biosensors are becoming small distributed biomedical systems rather than isolated transducers. Their practical performance depends on how the sensing interface, analog front end, power source, local computation, wireless link, packaging, and therapeutic output interact over time. The analysis focuses on that cross-layer problem, with emphasis on nanomaterial interfaces, energy autonomy, edge intelligence, and long-term reliability. Graphene, carbon nanotubes, MXenes, and transition-metal dichalcogenides are discussed across electrochemical, field-effect, impedance, optical, and radio-frequency transduction. Mechanical nanogenerators, biofuel cells, wireless power transfer, and hybrid storage are compared using the energy actually available after rectification and regulation rather than peak generator output alone. Quantitative re-plots illustrate non-monotonic carbon-nanotube loading in triboelectric layers and voltage-tunable few-layer-graphene microwave components. Edge AI is treated as part of the power and measurement architecture: local inference can reduce radio traffic and latency, but introduces model drift, uncertainty, and update requirements. The final sections connect biofouling, encapsulation, mechanical fatigue, calibration drift, wireless safety, and algorithm lifecycle to a common validation ladder for wearable, insertable, and implantable systems.

Bioengineering doi: 10.3390/bioengineering13101128

Authors: Yuan Tao Pengyu Ren Peng Han

Auditory brainstem response (ABR) threshold determination requires clinicians to identify the transition between waveforms with and without a detectable response across stimulus intensities. Automation remains challenging because responses near the threshold are weak, stimulus series are often incomplete, and the interpretation of one waveform depends on the remaining series from the same ear. To address these challenges, we represent each ear as a masked stimulus intensity response graph and propose ABR-GraphDoseNet for automated ABR threshold estimation. The model encodes waveform morphology at each stimulus intensity, exchanges information across an ordered intensity graph, and models candidate threshold transitions with a decoder for adjacent nodes. A global threshold head further integrates evidence across the complete ear series. We evaluated 2064 ears from 1045 subjects using ten-fold cross-validation grouped by subject. ABR-GraphDoseNet achieved the best overall performance, with a mean absolute error of 0.90±0.27 dB, an exact accuracy of 93.80±1.41%, and an accuracy within 10 dB of 98.64±1.01%. Ablation experiments identified the global threshold head and morphology tokenizer as the largest contributors, while adaptive edges and the boundary decoder provided smaller complementary gains. Overall, ABR-GraphDoseNet provides an accurate, structured, and auditable framework for modeling incomplete ABR stimulus series. External validation across devices, acquisition protocols, and institutions remains necessary.

Bioengineering doi: 10.3390/bioengineering13101127

Authors: Wen Chen Mingzhen Xu Bingxin Gu Jianping Zhang Xiaoqiang Shi Ni Zhang Shaoli Song

Objective: To develop and validate a clinical–radiomics nomogram based on 18F-FDG PET/CT for differentiating lymphoma from cancer of unknown primary (CUP) in head and neck metastatic poorly differentiated cancer. Methods: This retrospective study enrolled 192 patients, including 151 CUP and 41 lymphoma cases. Patients were randomly divided into a training (n = 134, 70%) and test cohorts (n = 58, 30%) using a fixed random seed for partitioning. Radiomic features were separately extracted from PET/CT images, and feature selection was performed using the least absolute shrinkage and selection operator algorithm. The optimal radiomics model was constructed via ten-fold cross-validation and comparative modeling with multiple machine learning methods. The clinical–radiomics nomogram was established by combining the radiomics model score and the clinical model score derived from gender and the metabolic heterogeneity index. The class distribution was preserved to reflect real-world clinical prevalence, consistent with recommendations for clinical prediction models. Results: The lymphoma group had a significantly higher female proportion (53.7%) than the CUP group (17.9%). The radiomics model yielded areas under the curve (AUC) of 0.916 in the training cohort and 0.812 in the test cohort. The final clinical–radiomics nomogram achieved higher and more robust performance, with AUCs of 0.924 (95% CI: 0.868–0.980) in the training set and 0.912 (95% CI: 0.837–0.987) in the test set, respectively. The standalone radiomics model exhibited a drop in sensitivity when evaluated in the test cohort, which directly demonstrates the incremental predictive value of integrating well-selected clinical parameters into the final model. Conclusions: The clinical–radiomics nomogram shows promising potential for differentiating lymphoma from CUP in poorly differentiated cervical lymph node malignancies. As a preliminary, single-center study, these findings warrant further validation in larger, multi-center cohorts before clinical implementation.

Bioengineering doi: 10.3390/bioengineering13101126

Authors: Sameer Ahmad Hasan Bassam Al-Naami Muhammad Al-Ayyad Abdel-Razzak Al-Hinnawi Olga Korostynska

Interest in local warming for chronic wound care has re-emerged as flexible heaters, embedded sensors, and smart dressings make controlled heat delivery increasingly feasible. This narrative translational review treats thermotherapy as a dose-dependent wound-interface intervention rather than a single warming technology. It distinguishes normothermic radiant warming, water-filtered infrared A, resistive Joule heating, photothermal systems, and temperature-monitoring technologies by mechanism, thermal context, therapeutic purpose, and evidence level. The review integrates wound pathophysiology, biological responses to temperature, thermal-dose concepts, emerging heater platforms, sensing and closed-loop control, clinical evidence, safety, and regulatory translation. Direct human treatment evidence remains concentrated in small historical radiant-warming and water-filtered infrared A studies, with limited additional local-heat evidence, and cannot be transferred directly to contemporary flexible, wireless, or photothermal platforms. For lower-temperature warming, heater setpoint alone is an inadequate dose descriptor, and the biological meaning of cumulative equivalent minutes at 43 °C (CEM43) has not been validated; measured time–temperature profiles, peak temperature, spatial variation, sensor location, and uncertainty are more informative. Clinical translation therefore depends on reproducible wound-interface dosing under realistic moisture, pressure, perfusion, dressing, motion, and workflow conditions, together with wound-specific dose models, transparent reporting, realistic interface validation, and architecture-specific regulatory evidence.

Bioengineering doi: 10.3390/bioengineering13101125

Authors: Chul Hee Jung Seok Yeon Choi Dong Ha Lee

Background: Interferential current therapy (ICT) and spray-and-stretch are two low-cost, non-invasive modalities that coexist on the treatment menus of outpatient musculoskeletal clinics, yet no published study has directly compared them, and no guidance document provides principled criteria for choosing between them. Objective: To characterise the mechanistic basis, clinical evidence, safety profile, and practical implementation of each modality across the full range of musculoskeletal disorders, and to propose a hypothesis-generating framework for modality positioning in the absence of direct comparative trials. Methods: A narrative review of literature retrieved from PubMed/MEDLINE, Embase, Scopus, the Cochrane Library, PEDro, and Google Scholar from database inception to 1 August 2026, using the search logic (musculoskeletal condition terms) AND (ICT terms OR spray-and-stretch terms), supplemented by hand-searching. The combined search retrieved 3241 records across six databases; 46 primary studies were included after two-reviewer sequential screening. Risk of bias was assessed using the PEDro scale (RCTs) and the Downs and Black checklist (observational studies). No quantitative synthesis was performed. Results: For ICT, multiple randomised trials and meta-analyses across at least four anatomical regions show that stand-alone ICT produces a small and inconsistent analgesic effect versus sham, whereas ICT as a co-intervention yields modest short-term improvements in pain and disability in chronic non-specific low back pain and knee osteoarthritis. For spray-and-stretch, a small number of predominantly fair-quality RCTs, almost all limited to the upper trapezius, show a real but brief effect of the combined procedure; the independent contribution of the vapocoolant has not been isolated in any dismantling study. Conclusions: The two modalities address mechanistically distinct clinical problems and are not directly comparable on the current evidence. We propose a hypothesis-generating positioning framework—unvalidated and intended to guide future comparative trials rather than to serve as a clinical decision rule—in which ICT is considered for deep, regional, multi-session analgesic adjunction, and spray-and-stretch for single-session restoration of range of motion around a discrete taut band. Direct comparative trial data are needed to test this framework.

Bioengineering doi: 10.3390/bioengineering13101124

Authors: Wan Azani Mustafa Norhayati Mohd Zain Halimaton Hakimi Muhammad Naufal Mansor Debrina Puspita Andriani

Artificial Intelligence (AI) has been created to be a potent tool that can help clinical decision-making by offering objective and ongoing monitoring of neonatal health. Unlike adults, infants are unable to communicate pain verbally and conventional pain measurement relies primarily on subjective and observational pain scales. This dependence introduces the possibility of some interrater variability and observer bias that may result in variability in assessment and suboptimal pain management in neonatal intensive care settings. To address these challenges, this study aims to provide a complete overview of the literature on AI-based infant pain detection and draw an intellectual map of this field by conducting a comprehensive bibliometric and science-mapping review. The systematic search was carried out by employing the Scopus database with an advanced and structured keyword search query to retrieve relevant English-language journal articles from 2010 to 2026. After screening with predefined inclusion criteria, a final set of 153 core documents was retained for analysis. The research productivity and citation impacts of the identified articles were assessed, and emerging thematic structures were identified through bibliometric processing and network visualization using a web-based bibliometric analysis platform, BiblioSpy®. Quantitative results demonstrate that the field is dynamic and expanding rapidly, with an annual publication growth rate of 29.06% and 2474 cumulative citations. The h-index of the analyzed corpus is 26, and the g-index is 46, with an average of 6.08 authors per publication and an 88.24% international co-authorship rate, suggesting a very collaborative environment. To conclude, quantitative keyword temporal tracking confirms this evolution: traditional binary algorithms display earlier average publication timelines (2019.0), whereas deep learning, Convolutional Neural Networks (CNNs), and multi-signal predictive models dominate recent outputs (2023.8–2024.5), demonstrating a quantitative shift from isolated binary classifiers toward continuous, explainable, and clinically viable multimodal monitoring systems. The results indicate that although the initial studies have been done on only binary classification of facial expressions and/or cry acoustics, the current research is moving towards continuous, explainable, and clinically viable multimodal monitoring systems.

Bioengineering doi: 10.3390/bioengineering13101123

Authors: Mia Rovis Alma Smajić Marijela Miličević Sandi Baressi Šegota Vedran Mrzljak Antun Gršković Juraj Ahel Klara Smolić Dean Markić Ivan Lorencin

This study evaluates the accuracy of medical-report interpretations generated by large language models in comparison to medically verified sources, with a particular focus on urology. The main goal is to examine the extent to which LLMs can reliably and precisely explain medical findings, with emphasis on expert urological terminology. For the evaluation of interpretation accuracy, real clinical documentation was used, specifically a dataset of 40 urological reports from the Clinical Hospital Center Rijeka. The generated explanations were compared with medically correct definitions using quantitative metrics assessing interpretability, specificity, and clinical accuracy. The results revealed a clear stratification of model performance, with DeepSeek-V3.2, GLM-4.6, MiMo-V2-Flash, and Qwen2.5-7B-Instruct achieving the highest semantic accuracy and consistency, while several models exhibited unstable behavior characterized by occasional catastrophic failures. Overall, the findings indicate that although LLMs can produce clinically coherent explanations of urological findings, their variability and susceptibility to hallucinations necessitate human oversight, supporting their use primarily as decision-support tools rather than autonomous clinical interpreters.

Bioengineering doi: 10.3390/bioengineering13101122

Authors: Gongmin Rim Zhongsoo Lim Kwanyong Hyun

Background: Thrombus formation remains a major complication during extracorporeal membrane oxygenation (ECMO), potentially resulting in circuit failure and thromboembolic events. Although several techniques have been proposed for thrombus detection, the relationship between measured ultrasonic features and target size has not been systematically characterized. This study evaluated size-dependent ultrasonic signal changes using standardized cylindrical surrogate targets under controlled conditions. Methods: A through-transmission ultrasonic system incorporating paired point-focused transducers (5 and 10 MHz), a water chamber, an ultrasonic pulser/receiver, a digital storage oscilloscope, and a manual translation stage was constructed. Cylindrical targets made of acrylic, acrylonitrile butadiene styrene (ABS), and SUS304 stainless steel, with diameters ranging from 0.1 to 5.0 mm, were scanned at 0.5 mm intervals over a 30 mm range. Baseline-subtracted waveforms were used to calculate peak height, pulse area, and squared-amplitude sum. Results: Exploratory regression analyses demonstrated positive diameter-dependent associations for all evaluated features. Pulse area showed strong linear associations across the six material–frequency conditions (R2 = 0.904–0.997), although no single feature consistently demonstrated the highest goodness of fit. Strong associations were observed at both frequencies; however, the different receiver gains precluded direct inference regarding relative frequency sensitivity. Under the 10 MHz condition, the 0.3 mm SUS304 wire target was the smallest tested surrogate target meeting the study-specific operational detection-index criterion (DI > 3). Conclusions: Through-transmission ultrasound demonstrated size-dependent changes in signals obtained from standardized cylindrical surrogate targets. Pulse area was identified as a practical candidate feature because of its consistently strong associations and computational simplicity, but neither statistical superiority nor a predictive sizing model was established. The 0.3 mm result applies to a high-acoustic-contrast SUS304 target under the specified experimental conditions and does not represent a biological thrombus detection limit.

Bioengineering doi: 10.3390/bioengineering13101121

Authors: Selma Dervisbegovic Verena Arnold Oleh Andrukhov Alexandra Wolf Apostolos Georgopoulos Hassan Shokoohi-Tabrizi Xiaohui Rausch-Fan Andreas Moritz Markus Laky

Persistent Enterococcus faecalis infection in inaccessible dentinal regions contributes to endodontic treatment failure, and adjunctive diode laser disinfection has been proposed to reach these areas, but direct comparisons of 445 nm and 970 nm wavelengths in a standardized lateral canal model are lacking. The aim of this study was to evaluate the antibacterial effect of a diode laser through a 0.5 mm dentin barrier in a clinically relevant ex vivo model and to compare its efficacy in reducing Enterococcus faecalis at wavelengths of 445 nm and 970 nm. Fifteen extracted human anterior teeth were instrumented to ProTaper F3, embedded in acrylic resin, and prepared with a standardized lateral canal 0.5 mm from the main canal. Following steam sterilization, the canals of specimens were inoculated with E. faecalis. Each tooth was exposed to all three conditions (control, 970 nm, 445 nm) across six successive experimental runs, and laser irradiation was performed through root dentin using a SiroLaser Blue diode laser (Dentsply Sirona), yielding 26 run-level measurements per condition overall. For each tooth, log10-transformed CFU/mL values were averaged within each condition prior to statistical analysis, giving one paired value per tooth per condition (n = 15). The mean log10-transformed bacterial load was 4.77 ± 0.12 in the control condition, 4.54 ± 0.11 following 970 nm irradiation, and 4.06 ± 0.75 following 445 nm irradiation. The Friedman test showed a significant overall difference among conditions (χ2(2) = 15.60, p < 0.001; Kendall’s W = 0.52), and Holm-adjusted Wilcoxon signed-rank pairwise comparisons (two-sided α = 0.05) showed significantly lower bacterial recovery following both 445 nm (mean tooth-level relative difference 55.9%; adjusted p = 0.0013) and 970 nm irradiation (38.2%; adjusted p = 0.0013) compared with the control, with significantly lower recovery following 445 nm than 970 nm irradiation (32.1%; adjusted p = 0.0215); these relative differences varied considerably among individual teeth. These findings indicate that diode laser irradiation exerts a wavelength-dependent antibacterial effect beyond the directly irradiated canal, supporting its potential as an adjunct to conventional root canal disinfection.

Bioengineering doi: 10.3390/bioengineering13101120

Authors: Kamil Çınar Christopher Mayack Hülya Eraslan-Gültekin Ali Fuat Ergenç Melike Kefeli Asiye Gök Yurttaş

Nosemosis, caused by the microsporidian parasites Nosema apis and Nosema ceranae, is a major contributor to honeybee (Apis mellifera) colony losses, yet its diagnosis still relies on destructive sampling followed by light microscopy or nucleic-acid amplification in a laboratory. We describe an impedimetric transducer intended as a first step toward a non-destructive alternative. Interdigitated gold electrodes were patterned on standard microscope slides by nanosecond laser ablation, a route that requires neither photolithography nor cleanroom processing, and were functionalized with (3-aminopropyl)triethoxysilane (APTES) deposited from either ethanol or toluene. Because the 0.01–50 MHz measurement window lies far above the range in which faradaic charge transfer contributes measurably, the response is non-faradaic and is governed by the permittivity and the ionic conductivity of the deposited droplet rather than by interfacial charge transfer. Films deposited from ethanol gave the largest and most reproducible contrast between an aqueous spore suspension (1500 spores/µL) and a deionized-water control. At the structural resonance of the electrode (≈16 MHz) the contrast in impedance magnitude is approximately 10% and is not resolved above the chip-to-chip scatter, which is expected because the magnitude at resonance is set by the series resistance of the electrode path rather than by the droplet. The two liquids are instead separated by the reactive component: the imaginary part of the impedance differs by 15.7 ± 1.8 MΩ near 6.8 MHz, about nine times the replicate uncertainty, and the phase by 33 ± 7° at 4.5 MHz. A self-contained readout built around an ESP32 microcontroller reproduced the direction of this change, its output voltage rising by approximately 33% after the suspension was applied. The scope of these observations is deliberately narrow: a single spore concentration in deionized water, no spore-free filtrate control, no calibration series, no selectivity panel and no measurement in a fecal matrix. The contribution is accordingly methodological—a cleanroom-free fabrication route, a quantified solvent dependence of the APTES film, and the transfer of the measurement from a benchtop spectrometer to a portable unit—rather than a validated diagnostic assay.

Bioengineering doi: 10.3390/bioengineering13101119

Authors: Raphaël F. Canadas

Cancer continues to impose a major global health burden, with an estimated 20 [...]

Bioengineering doi: 10.3390/bioengineering13101118

Authors: Mert Basancelebi Irem Senyer Yapici Rukiye Uzun Arslan

Routine complete blood count (CBC) measurements are widely available and may support the initial assessment of thalassemia-related hematological conditions. However, overlapping hematological profiles and class imbalance can complicate differentiation among groups. This study evaluated seven supervised machine learning classifiers for multiclass classification of normal hematological profiles, suspected α-thalassemia profiles, Hb E carrier-related profiles, and other thalassemia-related/hemoglobinopathic profiles using a publicly available CBC dataset. An unresampled baseline was compared with four synthetic oversampling strategies: SMOTE, Borderline-SMOTE, SVM-SMOTE, and KMeans-SMOTE. Model development and hyperparameter optimization were performed within five-fold stratified cross-validation, with model selection based on cross-validated Matthews correlation coefficient (MCC). The untouched internal hold-out test set was reserved for final evaluation. CatBoost without resampling yielded the highest cross-validated MCC (0.6378). On the internal hold-out test set, it achieved an accuracy of 0.7950, an MCC of 0.6537, Cohen’s kappa of 0.6499, and micro- and macro-average ROC–AUC values of 0.9554 and 0.9227, respectively. However, class-specific sensitivity was heterogeneous, ranging from 0.4028 for suspected α-thalassemia to 0.9767 for the normal hematological profile. Synthetic oversampling did not consistently increase overall CV MCC, although sensitivity improved for some minority classes. These findings characterize the performance and limitations of CBC-based multiclass classification and indicate the need for further external and clinically oriented validation before screening or referral applications can be considered.

Bioengineering doi: 10.3390/bioengineering13101117

Authors: Se Yeong Jeon Chul Hee Jung Seok Yeon Choi Dong Ha Lee

Insoles are among the most widely prescribed devices in orthopaedic practice, spanning conditions as mechanically dissimilar as paediatric flexible flatfoot, painful pes cavus, diabetic neuropathy, inflammatory arthritis, medial knee osteoarthritis, limb-length discrepancy, and neuromuscular gait disorders. Despite this breadth, the field is characterised by a persistent and instructive paradox: the biomechanical effects of insoles are reproducible and quantifiable, whereas their symptomatic effects are inconsistent and frequently indistinguishable from sham. This narrative review, drawing on a bounded, non-systematic evidence base of 55 sources across ten conditions, argues that this paradox is at least partly an engineering and measurement problem rather than a purely clinical one. We develop a mechanistic taxonomy in which every insole acts through some combination of four pathways—plantar pressure redistribution, external joint moment modulation, soft-tissue strain reduction, and sensorimotor modulation—and map the controllable design parameters (arch support height, medial posting angle, heel cup depth, thickness, material stiffness, and lattice architecture) onto each pathway using sensitivity data from a small number of heterogeneous finite element and Taguchi-based optimisation studies, several of which are single-model or single-patient analyses with limited experimental validation. We then review the manufacturing pathway from morphological capture through computer-aided design, additive or subtractive fabrication, and in-shoe verification, and summarise the condition-specific clinical evidence across the ten indications. Three hypothesis-generating findings emerge, each requiring prospective confirmation before being treated as established. First, custom devices consistently modify mechanics more than prefabricated devices, yet consistently fail to outperform them on symptom endpoints, indicating that the mechanism–outcome link is mediated by unmeasured variables. Second, the biomechanical response is heterogeneous in several conditions—for example, roughly one-third of patients with medial knee osteoarthritis show an increase rather than a decrease in the knee adduction moment with a lateral wedge—raising the possibility that population-averaged trials may obscure responder-specific effects, although the one trial that acted on this logic by individually tailoring the wedge angle found no clinical benefit. Third, a secondary analysis of a single randomised trial in painful pes cavus found that the redistribution, rather than magnitude, of plantar pressure predicted pain relief; if replicated, this would suggest that peak plantar pressure is an incomplete outcome measure in that specific pain-related context, a claim we do not extend to conditions such as diabetic offloading, where peak local pressure remains a mechanistically appropriate proxy for a different, site-specific outcome. We propose a design-decision framework, presented explicitly as a falsifiable hypothesis rather than a validated protocol, that selects design parameters by target mechanism rather than by diagnostic label, and argue that closed-loop prescription—instrumented insoles providing the continuous verification of both mechanical target attainment and wear adherence—is a promising but as-yet-unproven direction for the field.

Bioengineering doi: 10.3390/bioengineering13101116

Authors: Keita Nishi Daisuke Endo Takashi Hasegawa Shinichi Matsumoto Masaru Itose Madoka Makino Kazunobu Saiki Toshiyuki Tsurumoto Keiko Ogami-Takamura

Lumbar lordosis influences load transmission in the lumbar spine, but its effects on stress in the anterior cortical bone of the vertebral bodies remain unclear. This study investigated how lumbar lordosis angle (LL) affects vertebral stress. A finite element model from T12 to the sacrum was constructed from computed tomography images of a cadaver from a 66-year-old male donor. Three models with identical bone geometry and material properties but LL values of 50°, 30°, and 15° were analyzed under a 300 N compressive load and the same load combined with a 5 N·m flexion moment. Mean von Mises stress and the 95th percentile of von Mises stress were evaluated in the anterior cortical bone of L1–L5, and vertebral-level distributions were normalized to L1. Under both loading conditions, the LL15° model showed higher von Mises stress than the LL50° and LL30° models at all vertebral levels. Adding the flexion moment further increased stress in all models. Normalized stress was lowest at L3 in the LL50° and LL30° models, whereas values in the LL15° model remained relatively close to L1 across L2–L5. These findings suggest that models with lower LL exhibit higher anterior cortical bone stress and altered vertebral-level stress distributions.

Bioengineering doi: 10.3390/bioengineering13101115

Authors: Bernardo G. Collaco Nadia G. Wood Yunguo Yu Carina Rosa Malena Srinivasagam Prabha Ashton L. Boon Zhihui Fang Anjali Bhagra Bradley C. Leibovich Antonio Jorge Forte

Background: Large language models (LLMs) are increasingly used in patient-facing clinical applications, creating a need for scalable methods to evaluate the accuracy, safety, and appropriateness of their responses. Although physician review remains the reference standard, it is resource-intensive and difficult to scale. Objective: To develop and clinically validate an LLM-as-a-judge framework for evaluating responses to patients’ periprocedural questions. Methods: We retrospectively evaluated 1285 patient–AIVA interactions from 112 patients across two Mayo Clinic sites and multiple surgical specialties. Physician reviewers classified interactions using a predefined true-positive (TP), false-negative (FN), true-negative (TN), and false-positive (FP) framework. The finalized LLM judge independently evaluated the same interactions. Agreement was assessed using four-class and category-specific agreement, Cohen’s kappa, and a secondary binary analysis of response correctness. Results: Overall, four-class agreement was 93.1% (95% CI, 91.7–94.5%), with Cohen’s κ = 0.852 (95% CI, 0.823–0.882). Category-specific agreement was 94.1% for TP, 90.7% for FN, 95.5% for TN, and 84.9% for FP classifications. In the binary correctness analysis, accuracy was 93.4% (95% CI, 92.0–94.7%), sensitivity 94.6%, specificity 90.2%, precision 96.2%, and F1-score 0.954. McNemar’s test demonstrated no significant asymmetry between paired classifications (p = 0.11). Conclusions: A clinically grounded LLM judge closely approximated physician evaluation of patient-facing AI responses. These findings support its potential as a scalable assistive monitoring tool while preserving physician oversight for uncertain or safety-sensitive interactions.

Bioengineering doi: 10.3390/bioengineering13101113

Authors: Thong Nguyen Carter Woods Jian Liu Chris Wang Ai-Ling Lin Jianlin Cheng

The apolipoprotein E ε4 (APOE4) allele is the strongest genetic risk factor for late-onset Alzheimer’s disease (AD), the most common form of dementia. APOE4 carriers exhibit cerebrovascular and metabolic dysfunction, structural brain alterations, and gut microbiome changes decades before the onset of clinical symptoms. Better understanding of the early manifestation of these physiological changes is critical for development of timely AD interventions and risk reduction protocols. Multi-modal datasets encompassing a wide range of APOE ε4 and AD associated biomarkers provide a valuable opportunity to gain insight into the APOE4 phenotype; however, these datasets often present analytical challenges due to small sample sizes and high heterogeneity. Here, we propose a two-stage multimodal AI model (APOEFormer) that integrates blood metabolites, brain vascular and structural MRI, microbiome profiles, and other clinical and demographic data to predict APOE4 allele status. In the first stage, modality-specific encoders are used to generate initial representations of input data modalities, which are aligned in a shared latent space via self-supervised contrastive learning during pretraining. The contrastive learning objective encourages learning of informative and consistent representations across modalities through leveraging cross-modality relationships. In the second stage, the pretrained representations are used as inputs to a multimodal transformer that integrates information across modalities to predict a key AD-risk genetic variant (APOE4). Across 10 repeated experimental runs with different train–validation–test splits, APOEFormer predicts whether an individual carries an APOE4 allele with an average prediction accuracy of 75%, demonstrating robust performance under limited sample sizes. Post hoc perturbation analysis of the predictive model revealed valuable insights into the driving components of the APOE4 phenotype—including key blood biomarkers and brain regions strongly associated with APOE4.

Bioengineering doi: 10.3390/bioengineering13101114

Authors: Daniela Hartmann Aimée Braun Benjamin Kendziora Katharina Wex Paulina Pabst Andreas Ohlmann Elke C. Sattler Maximilian Deußing

Cutaneous T-cell lymphoma (CTCL) often clinically resembles inflammatory dermatoses such as eczema, making accurate and timely diagnosis challenging. Since early differentiation is critical for appropriate treatment and prognosis, fast and reliable diagnostic tools are needed. Ex vivo confocal laser scanning microscopy (EVCM) has emerged as a promising technique for rapid, high-resolution tissue assessment. As CTCL is difficult to identify based on clinical presentation alone and is frequently misdiagnosed as eczema, this study assessed the diagnostic accuracy and clinical utility of EVCM in differentiating CTCL from eczema, as well as interobserver agreement among examiners with varying experience. Fifty specimens (25 eczema, 25 CTCL) were imaged using EVCM (Vivascope 2500 M-G4; Vivascope GmbH, Munich, Germany). Three independent, blinded dermatologists with different experience levels classified lesions based solely on digital H&E-like images. Sensitivity, specificity, and predictive values were calculated. EVCM showed high diagnostic accuracy, with a mean concordance rate of 82.67%. The dermatohistopathologist achieved the highest accuracy (84%), followed by two trained dermatologists (82% each). Sensitivity ranged from 0.80 to 0.88 and specificity from 0.76 to 0.84, with comparable predictive values and almost perfect interobserver agreement (Fleiss’ κ = 0.89). Misclassifications mainly involved the same diagnostically challenging cases across both entities. EVCM demonstrated high diagnostic accuracy across examiners of varying experience, supporting its potential as a rapid bedside tool for distinguishing CTCL from eczema. While not replacing histopathology as the gold standard, it may meaningfully complement routine diagnostics. Integration of immunohistochemistry or artificial intelligence could further enhance diagnostic precision.

Bioengineering doi: 10.3390/bioengineering13101112

Authors: Bingyu Pan Yexuan Wang Mingzi Xiang

Surface electromyography (sEMG) reflects neuromuscular control, but multichannel recordings are high-dimensional and difficult to interpret. This study evaluated whether muscle-synergy activations provide a compact input representation for lower-limb kinematic reconstruction. Twelve-channel sEMG and hip, knee, and ankle angles were obtained from 120 healthy male participants performing seven tasks in the Gait120 dataset. A four-synergy representation was derived using nonnegative matrix factorization, and within-participant cross-task similarity was assessed. Under participant-wise five-fold cross-validation, fold-specific shared dictionaries were estimated exclusively from training participants, and separate task-specific models reconstructed joint trajectories across five locomotor tasks. ExtraTrees achieved the lowest RMSE in most task–joint combinations and was used for exploratory detailed comparisons. Shared-synergy activations reduced the regressor-input representation from 12 variables to four activation coefficients per time point. Task-level mean RMSEs were 5.60–6.29° for synergy activations and 5.52–6.28° for raw sEMG. After Holm correction, no statistically significant difference was detected in 14 of the 15 task–joint comparisons, while one favored synergy activations. These findings support shared-synergy activation as a compact and physiologically interpretable input representation and provide a basis for neuromuscularly informed modeling of lower-limb movement across locomotor conditions.

Bioengineering doi: 10.3390/bioengineering13101111

Authors: Youzhuo Quan Huaizhi Yue Xiangyang Ma Kun Liu

Accurate segmentation of coronary vessels and stenotic regions in X-ray coronary angiography remains challenging because of low vessel-to-background contrast, vessel overlap, thin branches, and the small, ambiguous appearance of stenotic lesions. In this study, we propose DCD-VMUNet, a structure-aware framework for automated coronary vessel and stenosis segmentation. Built on VM-UNet, the model incorporates a dynamic multi-scale kernel module, a centerline head, and a direction head. The dynamic module adaptively integrates fine boundary details with broader contextual cues, while the auxiliary heads introduce geometric supervision to enhance structural continuity and orientation awareness. DCD-VMUNet was independently trained and evaluated for binary vessel and stenosis segmentation on the ARCADE dataset. It was compared with YOLOv11-X, DeepLab v3+, UNet, UNet++, TransUNet, MALUNet, and VM-UNet using pixel-wise metrics and, for vessel segmentation, the centerline-aware clDice metric. For vessel segmentation, DCD-VMUNet achieved the best performance, with IoU, precision, recall, and F1-score of 0.6838, 0.8418, 0.7846, and 0.8122. For stenosis segmentation, it also ranked first across all metrics, reaching 0.4042, 0.5826, 0.5688, and 0.5757, respectively. Ablation studies confirmed the complementary effects of multi-scale feature extraction and geometric supervision. Additional evaluation on the independent ICA dataset further supported the vessel segmentation performance of DCD-VMUNet beyond the ARCADE benchmark. These results indicate that DCD-VMUNet improves coronary vessel delineation and pixel-level stenosis localization, supporting automated coronary angiographic assessment.

Bioengineering doi: 10.3390/bioengineering13101110

Authors: Sheng Jin Kun Deng Haoqi Cai Yuchan Li Zhigang Wang Zhu Liu Yuling Li

Pediatric aneurysmal bone cysts (ABCs) are locally aggressive lesions with variable postoperative behavior, and quantitative imaging markers associated with healing or recurrence remain limited. This exploratory, retrospective, single-center study evaluated whether the relative transverse cyst index (RTCI), lesion-to-physis distance, and mean intralesional computed tomography (CT) attenuation were associated with radiographic outcomes after surgery. Twenty-four children with biopsy-confirmed primary ABCs treated between 2017 and 2023 underwent blinded dual-observer measurement of lesion dimensions, involved-bone length, physeal distance, and CT attenuation; outcomes were classified as complete healing, recurrence, or stable partial residual disease. RTCI differed across outcome groups (0.069 ± 0.024, 0.126 ± 0.046, and 0.114 ± 0.0534, respectively; p = 0.019, false-discovery-rate-adjusted q = 0.045), as did lesion-to-physis distance (4.58 ± 3.64, 1.04 ± 1.09, and 1.26 ± 1.95 cm; p = 0.023, q = 0.045). Mean CT attenuation did not differ among groups (p = 0.843). Greater relative transverse expansion and closer physeal proximity were associated with less favorable radiographic outcomes. These hypothesis-generating findings require validation in larger multicenter cohorts.

Bioengineering doi: 10.3390/bioengineering13101109

Authors: Abdulaziz Alshehri Abdulrahman Alshabib Yasser Alfawaz Abdulaziz Alayed Nick Silikas Mohammed Alrabiah Silvia Rojas-Rueda Hamad Algamaiah

Background: This study evaluated long-term water sorption, solubility, and hygroscopic expansion of bulk-fill resin composites (BRCs), with and without fiber reinforcement, at 37 °C, focusing on the influence of material composition on stability in aqueous environments. Methods: Five resin composites were tested: fiber-reinforced everX Flow (EVX); particle-reinforced SDR, Filtek Bulk-fill Flowable Restorative (FBF), and Filtek Bulk-fill Posterior Restorative (FBP); and conventional Filtek Z350 XT (FZ). Nine disk-shaped specimens per material were evaluated for water sorption and solubility according to ISO 4049. Hygroscopic dimensional changes were measured using laser scanning. Filler mass and degree of conversion were also assessed. Results: After 80 days, water sorption ranged from 13.79 to 30.13 μg/mm3. FZ showed the highest sorption and EVX the lowest. Flowable bulk-fill materials had lower sorption than packable materials. Solubility ranged from −3.10 to 0.73 μg/mm3. Hygroscopic expansion ranged from 0.64% to 1.51%, with EVX showing significantly lower expansion and FZ the highest. Conclusions: Short E-glass fiber-reinforced flowable composite showed lower hygroscopic expansion, yielding superior dimensional stability. Additionally, material solubility was highly matrix-dependent, with TEGDMA-free formulations demonstrating the lowest values. Conversely, applying the conventional composite beyond recommended increment depths severely impaired monomer conversion, which directly correlated with maximal water sorption and hygroscopic expansion.

Bioengineering doi: 10.3390/bioengineering13101107

Authors: Xiu Chen Baowen Zhang Mengdie Li Song Wang Mei Yang Jiejing Zhang Xuying Zhu Yingjie Kang Wenli Tan

Existing MRI-based prediction models for cognitive impairment often rely on a single biomarker or simply compare algorithms, act as black boxes, and lack external validation, which limits their clinical translation. We developed and externally validated an interpretable multimodal machine-learning model that classifies a MoCA-defined screening outcome as a binary task: cognitive impairment (CI, MoCA < 26) versus no cognitive impairment (NCI, MoCA ≥ 26). Sixteen candidate features were considered, including AI-derived medial temporal atrophy (MTA), regional white matter hyperintensity (WMH) volumes and proportions, and Fazekas ratings from routine T1WI and T2-FLAIR, together with age and sex. In a single-center cohort of 433 subjects (302 CI, 131 NCI) with a 7:3 training/hold-out split, five algorithms were compared and Random Forest performed best; SHAP-guided recursive feature elimination within five-fold cross-validation then reduced the set from 16 to 11 features. The optimized 11-feature Random Forest reached a cross-validated AUC of 0.883 ± 0.024 on the training folds and an internal hold-out AUC of 0.966, and 0.860 in an independent ADNI cohort of 139 subjects (100 CI, 39 NCI). SHAP identified the left MTA score as the most influential feature, followed by near-ventricular (juxtaventricular and periventricular) WMH burden and the right MTA score. The left MTA score was most strongly correlated with MoCA (r = −0.481). The model may offer an objective, routine-MRI-based screening aid, pending prospective multicenter validation.

Bioengineering doi: 10.3390/bioengineering13101108

Authors: Almudena Banegas-Cano Arturo Sánchez-Pérez Alfonso Jornet-García José María Montoya-Carralero María José Moya-Villaescusa

Immediate implant mechanics may depend on the interaction between implant macrogeometry and osteotomy preparation, but their combined influence on complementary measures of immediate fixation and regional thread alterations remains insufficiently characterized. This ex vivo study evaluated two implant macrogeometries and four osteotomy techniques in 80 bovine rib specimens (n = 10 per group). Reverse torque resistance remained the primary mechanical outcome; prospectively recorded insertion torque and resonance frequency analysis (implant stability quotient [ISQ]) were incorporated in the revision as complementary immediate mechanical outcomes. Thread alterations were assessed qualitatively after removal in the cervical, body, and apical regions. Insertion torque was significantly influenced by macrogeometry and preparation technique, without a significant interaction. Reverse torque resistance showed significant effects of macrogeometry, preparation technique, and their interaction (all p < 0.001). ISQ showed a significant macrogeometry-by-technique interaction. Insertion torque correlated strongly with reverse torque resistance (r = 0.645, p < 0.001), whereas ISQ showed a more moderate association (r = 0.370, p < 0.001), supporting related but non-interchangeable mechanical constructs. Cervical alterations were mainly associated with undersized drilling, whereas apical alterations occurred predominantly in the Quattro design; none were observed in the implant body. External surface temperature showed no overall change or consistent technique-dependent increase. Under these experimental conditions, implant macrogeometry and osteotomy preparation acted as interdependent determinants of immediate mechanical behavior and were associated with distinct regional patterns of post-removal, visually detectable thread alterations.

Bioengineering doi: 10.3390/bioengineering13101106

Authors: Zhonggang Feng Tetsuro Uchida

Tissue-engineered constructs (TECs) offer promising alternatives to autologous grafts and conventional artificial implants for the repair or replacement of damaged soft tissues. Nevertheless, their clinical translation remains limited for several reasons. While biomaterial biocompatibility, immune regulation, and the utilization of stem cells have received considerable attention, the mechanical environment of the graft–host junction (GHJ), particularly the mechanical mismatch between grafts and host tissues, has not been paid enough attention. This review proposes that mechanical continuity is as fundamental as biological compatibility for durable graft integration and examines the GHJ as a spatiotemporally evolving mechanobiological system coupled with immune surveillance and responses. In this review, we first draw lessons from two natural tissue transitions that involve changes in mechanical properties: the development of the muscle–tendon–bone unit and skin-wound healing processes. We then survey current clinical applications of soft-tissue TECs. It comes out that complications frequently arise at or near the graft–host junction; however, the limited clinical characterization of GHJ mechanics also indicates that interface mechanobiology remains substantially overlooked in TEC clinical applications. In view of this situation, we analyze the potentially critical roles of mechanobiology at the GHJ, characterizing graft integration as intertwined spatiotemporal processes governed by both general and mechano-immune coupling. We finally advocate for incorporating mechanobiological considerations into graft interface design to achieve functional integration in clinical TEC applications and highlight AI-assisted inverse design and multimaterial 3D bioprinting as the most promising technologies to realize mechanobiological design and fabrication of the TEC junction portion.

Bioengineering doi: 10.3390/bioengineering13101105

Authors: Beatrice De Lazzari Angela Richter Christoph Nix Roberto Badagliacca Annalisa Pitino Mercedes Gori Gianmarco Scoccia Massimo Capoccia Claudio De Lazzari

Background: Indications for right ventricular assist device (RVAD) insertion include right heart failure after implantation of a left ventricular assist device or early graft failure following heart transplantation. This study aimed to investigate how the upstream and downstream circulatory network interacts with the Impella RP® device. Methods: A numerical model of the Impella RP® device was implemented within the CARDIOSIM© software platform for this study. In the numerical configuration, the RVAD aspirated blood from either the right atrium (RA-PA connection) or the right ventricle (RV-PA connection) and delivered it to the pulmonary artery. Only RA-PA connection is the currently used setting for the Impella RP® device in clinical practice. Based on right ventricular (RV) decompression and total flow, our study may help define the need for a direct RV-unloading Impella RP®. Results: The simulations showed that while activating the RVAD in RA-PA mode, regardless of its rotational speed, the mean pulmonary artery pressure (PAP) percentage change was higher than that in the unsupported condition when the mean systemic venous pressure (SVP) and the pulmonary artery wedge pressure (PAWP) were both set to 20 [mmHg]. When RV-PA connection was applied, a similar trend was observed, although the PAP percentage changes were approximately halved compared to those when the RA-PA connection was applied. Conclusions: The Impella RP® device has the potential to become a valid option for RV support based on current experimental and simulation data. Although already in use, further evaluation in the clinical setting will likely confirm its potential and lead to a more routinely application for RV support.

Bioengineering doi: 10.3390/bioengineering13101104

Authors: Victor K. Tsui Kouhyar Tavakolian Daniel Ewert

Cardiovascular diseases are among the leading causes of death worldwide, creating a need for reliable hemodynamic testing platforms for cardiovascular device evaluation and therapeutic development. Mock circulatory loops (MCLs) provide effective in vitro platforms for reproducing physiological and pathological pressure conditions; however, many existing systems remain limited in programmability, waveform fidelity, standardization, and affordability. This study presents a cost-effective dual-pump hybrid MCL (hMCL) that independently regulates baseline and pulsatile pressure components using Fourier-based harmonic synthesis. Reference arterial pressure waveforms were decomposed into harmonic components, and their amplitudes, frequencies, and phase shifts were converted into pulse-width-modulation signals for dual-pump actuation. Validation using five physiological and virtual arterial pressure waveform datasets demonstrated MAP errors of ≤1.3%, RMSE values of 9.1–24.1 mmHg, and feature errors ≤ 2.8% for systolic, diastolic, and dicrotic notch pressures. Distribution analysis showed small median pressure differences and low variability across waveform characteristics, while repeated hMCL operation demonstrated pressure variations below 3 mmHg. These findings demonstrate that the proposed Fourier-based hMCL provides an accurate, reproducible, and cost-effective platform for arterial pressure waveform reproduction and early-stage cardiovascular device testing. The present validation is pressure-focused; flow waveform fidelity and pressure–flow interactions were not independently characterized.

Bioengineering doi: 10.3390/bioengineering13101103

Authors: Fatma Nur Kiliçkaya İşci Celal Öztürk Eda Köseoğlu Arzu Hanım Yay

Autophagy is essential for ovarian physiology and is implicated in pathological conditions such as Polyendocrine Metabolic Ovarian Syndrome (PMOS). The evaluation of autophagy-related markers in histopathological images is often subjective and observer-dependent. This study presents an explainable computational pathology framework that integrates patch-based deep learning with Grad-CAM-based visual interpretation to enable objective and reproducible analysis of autophagy-related immunohistochemical (IHC) staining in rat ovarian tissue. Beclin-1, LC3, and p62-stained rat ovarian tissue images from control, PMOS, bee bread (PERGA), and PERGA + PMOS groups were analyzed using a patch-based deep learning approach. Multiple convolutional neural network architectures, including DenseNet121, EfficientNetV2B0, and ConvNeXtTiny, were trained and evaluated, and their outputs were combined through an ensemble strategy. Image-level predictions were generated by averaging patch-level probabilities, and Grad-CAM was used to visualize image regions influencing model decisions. The LC3-based ensemble model achieved the highest overall image-level classification performance, with an F1-score of 78.28% and an AUC of 88.15%. In contrast, the Beclin-1 DenseNet121 model demonstrated the highest discrimination performance based on AUC, reaching 93.63%. Models trained on p62 images showed lower classification performance, likely due to the heterogeneous staining characteristics of p62 expression. Grad-CAM visualizations revealed that model activation was predominantly localized to DAB-positive and tissue-informative regions, supporting the interpretability of the proposed framework. In summary, this approach enables objective, reproducible, and interpretable analysis of autophagy-related histopathological images. The findings suggest that patch-based deep learning combined with explainable artificial intelligence offers a promising computational tool for assessing autophagy-related alterations in rat ovarian tissue.

Bioengineering doi: 10.3390/bioengineering13101102

Authors: Eric E. Poole Charles A. Odonkor Eric T. Nguyen David W. Lee Mustafa R. Dodurgali Emily Draeger Alaa Abd-Elsayed

Radiofrequency ablation (RFA) is a foundational interventional technology in pain management. Despite the diversity of available systems, conventional thermal, internally cooled, bipolar, and pulsed radiofrequency (PRF) all share one biophysical mechanism: high-frequency alternating current (AC) drives tissue ions to oscillate, generating frictional heat within the tissue surrounding the electrode. This localized heating produces predictable patterns of coagulative necrosis shaped by tissue electrical conductivity, electrode geometry, power delivery, and perfusion-related heat dissipation. Classical and contemporary bioheat models, including the Pennes bioheat equation, describe how these variables determine lesion size and morphology. Advances in device engineering have reshaped lesion characteristics through internally cooled probes, bipolar strip-lesioning systems, and pulsed platforms that modulate neural activity while preserving structural integrity. These innovations broaden therapeutic options by enabling targeting of anatomically variable structures, such as the sacral lateral branches, and intraosseous targets, including the basivertebral nerve (BVN). Emerging technologies—high-voltage PRF, hybrid pulsed–thermal systems, and electroporation-based pulsed-field ablation—reflect a growing emphasis on tissue selectivity, energy efficiency, and safety. This narrative review traces the historical development of radiofrequency neurotomy, examines the biophysical principles governing lesion formation (including the Pennes bioheat equation and the roles of current density, impedance, and perfusion), and details the engineering distinctions among modalities. Clinical applications across spinal, joint, and peripheral nerve targets are presented to illustrate how these physical and engineering differences translate into procedural decision-making, and emerging modalities such as high-intensity focused ultrasound (HIFU) and pulsed-field ablation are discussed in the context of the biophysical principles they extend or depart from.

Bioengineering doi: 10.3390/bioengineering13101099

Authors: Jung-Hyun Yun Jae-Myun Ko Tae-Hyung Lee

Commercial smartwatches are increasingly used to monitor physiological responses during exercise; however, their validity for resistance exercise remains uncertain. This study evaluated the criterion validity of four commercial smartwatches (Apple Watch Series 8, Samsung Galaxy Watch 5, Fitbit Charge 5, and Garmin Vivosmart 5) for measuring heart rate (HR) and energy expenditure (EE) during resistance exercise. Seventy healthy men aged 20–39 years performed two sets each of the bench press, T-bar row, squat, and deadlift at individualized 10-repetition maximum loads. Smartwatch-derived HR was evaluated against electrocardiography, whereas smartwatch-derived workout-session EE was evaluated against indirect calorimetry. The Apple Watch showed the lowest measurement error for mean HR across the protocol (bias = −0.36 bpm; MAE = 1.81 bpm; ICC = 0.969) and for overall dynamic HR (bias = −0.72 bpm; MAE = 4.95 bpm; ICC = 0.904). HR agreement deteriorated markedly immediately after each set, particularly for the Samsung and Fitbit devices, which underestimated ECG-derived HR by 25.60 and 19.59 bpm, respectively. In contrast, EE estimates showed poor agreement with indirect calorimetry across all devices (ICC = 0.021–0.033). These findings demonstrate that the agreement of smartwatch-derived HR with electrocardiography measurements varies across time points during resistance exercise and indicate that smartwatch-derived workout-session EE should be interpreted cautiously when accurate metabolic assessment is required.

Bioengineering doi: 10.3390/bioengineering13101101

Authors: Tzu-Wei Lin Guan-Jen Li Wei-Cheng Lien Yan-Tsung Peng Yi-Ting Lai Yu-Hsiu Chen

Parkinson’s disease (PD) is one of the most common neurodegenerative disorders in the aging population, which is commonly associated with motor abnormalities that can be reflected in hand-drawn geometric patterns. However, early detection remains challenging due to the lack of specific screening or tests, especially in rural areas with limited medical resources, leading to delays in medical intervention. To address these critical limitations, this paper proposes FusionNet, a dual-branch vision Transformer framework that combines spatial and frequency-domain information for handwriting-based PD screening. The spatial branch learns geometric and stroke-related representations directly from grayscale drawings, whereas the frequency branch processes log-scaled amplitude spectra obtained by two-dimensional fast Fourier transform. Features from the two branches are integrated using bidirectional cross-attention (BCA), allowing spatial and spectral representations to condition one another before classification. A staged training strategy was adopted in which the individual branches were trained separately before optimization of the fusion components. Experimental evaluations on the publicly available NewHandPD dataset were conducted across the three drawing tasks: hand-drawn circle, meander, and spiral images from 66 participants (31 individuals with PD and 35 healthy controls). Performance was evaluated using five-fold Group K-Fold cross-validation, demonstrating that FusionNet achieves the highest accuracy of 91.68% on the meander task and maintains >85% sensitivity with >88% specificity across all tasks. Because the study used a small single dataset, participant-level separation was not guaranteed for all tasks, and no independent external validation was performed, the reported results should be interpreted as preliminary and dataset-specific. Analysis of model behavior further indicates that spatial-domain predictions could be affected by paper texture, whereas the complementary frequency representation reduces this source of bias. Finally, the model was incorporated into a LINE-based prototype to demonstrate the technical feasibility of remote image submission and automated analysis. This implementation remains a proof of concept and requires external, prospective clinical validation before use in clinical decision-making.

Bioengineering doi: 10.3390/bioengineering13101100

Authors: Shuailin Du Guang Li Ren Togo Takahiro Ogawa Miki Haseyama

Medical image analysis often relies on large annotated datasets, whose collection, storage, and sharing are costly and constrained by privacy concerns. Dataset distillation offers a potential solution by constructing compact synthetic datasets that retain task-relevant information from the original data. However, generative distillation methods may not adequately preserve class-specific information in medical images, where clinically distinct classes often exhibit similar visual characteristics. To address this issue, we propose a classifier-derived guidance framework for diffusion-based medical dataset distillation. The framework uses predictions from a pretrained noise-conditioned classifier to guide diffusion sampling through three objectives. Classifier guidance (CG) promotes class alignment, whereas entropy and margin guidance encourage boundary-oriented sampling. We analyze guidance-strength sensitivity on three medical imaging datasets across images-per-class (IPC) budgets and evaluate validation-selected COVID-19 Radiography configurations on the test subset using three downstream architectures. Classifier diagnostics distinguish prediction–conditioning agreement from downstream utility, while learning curves illustrate differences in validation behavior despite low training loss. These findings characterize the configuration-dependent utility of classifier-derived guidance for compact synthetic medical datasets.

Bioengineering doi: 10.3390/bioengineering13101098

Authors: Tricia Morphew Hannah Pease Michelle A. Fortier Zeev N. Kain Lois W. Sayrs

Background: The growing use of computational modeling of real-world data (RWD) in clinical research introduces significant risks for data scientists and AI trainers already managing inconsistency, unexamined bias, and analytical opacity. Building upon more than 600 guideline-based checklists and pre-planning templates for data quality and transparency reporting in observational RWD, we propose a run-time operational execution loop to embed reproducibility analytics within the workflow. Our approach acts as a run-time gatekeeper mapping prespecified analytics goals to a closed-loop LLM semantic code auditor, checking for both logic, bias, and vulnerabilities during code execution. Our framework was designed to bridge the gap between passive disclosure, ensuring proposed requirements, and post-hoc validation. Our structured hierarchical approach coupled with AI auditing is the first phase in the development of full-stack automation of bias-aware RWD analytic validation. Methods: Our analytic framework spans 10 domains across three analytic phases. Each domain includes prespecified criteria governing verification with potential biases, mechanisms, and mitigation strategies systematically mapped throughout. Phase I entails specification (Domains 1–4: objectives and estimands, data inputs and outputs, cohort and filters, and variable coding); Phase II: modeling (Domains 5–7: model testing, diagnostics, and reasonableness checks); and Phase III: robustness and interpretation (Domains 8–10: missing data handling, temporal and stratified checks, and scope boundaries). We utilize specific prompt-directed large language model (LLM) auditing of model elements to ensure data quality. The methodology is illustrated with examples from an empirical characterization of inpatient opioid data from the Cerner Real-World Data (2014–2021). GLMMs and corresponding outputs were generated in R v4.4.2. Curated AI-auditing of first-pass code drafting was performed using ChatGPT 5.2 (OpenAI), with cross-LLM portability assessed using Claude Sonnet 5 (Anthropic). Results: The Cerner RWD database contained 16,734 pediatric cancer patients (76,395 encounters) across 55 U.S. health systems. During the specification phase, total daily dose could not be estimated with sufficient validity from the Cerner RWD, and “as-needed” (PRN) orders, warranted empirically supported exclusion. The modeling phase revealed residual variability in opioid ordering attributable primarily to patient-level differences within health systems (ICC = 19.9%) rather than between health systems (ICC = 2.0%). Temporal analysis revealed divergent opioid-ordering trajectories from 2017 onward, with decreasing morphine and increasing fentanyl ordering. We then utilized curated prompting to enforce an AI-generated coding requirement to produce valid likelihood ratio tests across the prespecified model sequence accounting for temporal trends. After harmonizing the model specification and estimation settings, both LLM workflows produced identical random-effect variance estimates, ICCs, and AIC values evaluated through the third decimal place, with corresponding fixed-effect estimates likewise matching. Conclusions: The 10-domain framework provides a structured, reproducible mechanism for governing RWD analyses by using a closed-loop semantic code auditor to embed run-time analytics directly into the data science workflow, supporting transparent exposure definition, systematic bias assessment, and prespecified analytic decision-making. By bridging the gap between passive transparency checklists and active automation, the framework operationalizes existing data quality guidelines through an executable, bias-aware code-auditing process. Cross-LLM concordance in identifying problematic logic and reproducing complex workflows supports the portability of this run-time gatekeeper. Ultimately, this approach establishes a scalable pathway for embedding expert statistical decision-making into the automation of safe, reliable real-world evidence (RWE) generation.

Bioengineering doi: 10.3390/bioengineering13101097

Authors: Srinivasagam Prabha Bernardo G. Collaco Carina Rosa Malena Syed Ali Haider Nadia Wood Antonio Jorge Forte

Background: Reliable access to postoperative guidance between clinic visits remains inconsistent, and the usefulness of large language models depends on how clinical knowledge is structured and retrieved. Knowledge graph-only (KG-only), vector retrieval-augmented generation (RAG-only), and hybrid graph-retrieval (GraphRAG) architectures have not been directly compared under controlled conditions for postoperative care. Objective: It was to compare these three pipelines on a curated postoperative question-answering benchmark using automated grounding metrics and blinded clinician review. Methods: A benchmark of 250 expert-authored questions (200 in-scope, 20 emergency-intent, 30 out-of-scope) was evaluated on an identical knowledge corpus using Gemini 2.5 Flash for generation and Gemini Embedding-001 for retrieval. Three blinded clinicians rated all 750 outputs; the primary outcome was majority-consensus accuracy across all 250 questions. GraphRAG is a multi-component pipeline that also includes query expansion, a second retrieval stage, evidence fusion, and a monotonic RAG fallback, and is therefore evaluated as a complete hybrid system rather than as graph retrieval in isolation. Results: Overall accuracy was 96.8% (95% CI: 93.8–98.4) for GraphRAG, 94.8% (95% CI: 91.3–96.9) for RAG-only, and 56.0% (95% CI: 49.8–62.0) for KG-only. Both retrieval-augmented systems outperformed KG-only (adjusted p < 0.001). The GraphRAG versus RAG-only difference was 2.0 percentage points, arose from five discordant question pairs and did not reach the Bonferroni-adjusted threshold (exact McNemar p = 0.062; adjusted p = 0.188). The monotonic RAG fallback determined 44.8% of GraphRAG outputs (112 of 250 queries). GraphRAG showed higher context recall (0.851 versus 0.796; adjusted p < 0.001), whereas RAG-only showed higher context precision. Estimated hallucination risk, derived from an LLM-as-a-judge procedure using Gemini 2.5 Flash rather than independent clinical adjudication, was 0.108, 0.132, and 0.436, respectively; the GraphRAG versus RAG-only difference was not significant (adjusted p = 0.771). Inter-rater agreement was high (Fleiss’ kappa = 0.854). Conclusions: Retrieval augmentation, rather than the choice of retrieval architecture, accounted for most of the difference observed. The incremental advantage of GraphRAG over a well-implemented RAG-only system was small, was not statistically significant after correction and reflects the complete hybrid system rather than graph augmentation alone. Emergency-intent accuracy was 75.0% for both retrieval-augmented systems; the architecture is therefore suited to supervised clinical workflows with clinician review, not autonomous patient-facing triage. Prospective validation on authentic patient messages and formal component ablations are required.

Bioengineering doi: 10.3390/bioengineering13101096

Authors: Alexandre Dontschev Vivienne Woodtli Clemens Maria Schiestl Kathrin Neuhaus Philippe Abdel-Sayed Anthony de Buys Roessingh Christoph Moor Ueli Moehrlen Sophie Böttcher

Cultured epithelial autografts (CEAs) have been used for more than 40 years in the setting of extensive total body surface area (TBSA) burns. The development and clinical implementation of cultured epidermal autografts and, subsequently, combined cultured dermo-epidermal autografts (CDEAs) have provided an essential adjunctive treatment option. Over the past five decades, numerous studies have reported on their clinical application. CEAs are manufactured from autologous epidermal cells that are expanded in vitro and assembled into sheet grafts suitable for transplantation. While substantial literature exists regarding their clinical use, data specifically addressing treatment-related complications remain scarce. This systematic review aims to comprehensively synthesize the available evidence on sheet-type CEA/CDEA use in burn patients, with a particular focus on complications, tumorigenicity, and mortality. A systematic literature search was conducted in 2023 across five electronic databases (Cochrane on 31 May 2023, Embase on 31 May 2023, Ovid Medline on 31 May 2023, Scopus on 1 June 2023, and Web of Science on 1 June 2023) to identify studies reporting on the use of CEAs in burn patients. Search terms included combinations of “cultured,” “epithelial,” “autografts,” “take rate,” “mortality,” and “skin cancer.” Eligible studies were human clinical studies involving burn patients treated with sheet-type CEA or CDEA, published in English, French, or German. Exclusion criteria comprised studies involving allogeneic materials, in vitro or animal studies, and commentaries and editorials. Efficacy endpoints included complications, tumorigenicity, and mortality outcomes in CEA-treated patients. Risk of bias was assessed for each included study using the modified Downs and Black checklist and graded according to Hooper et al. Seventy studies contributing 72 patient cohorts published between 1984 and 2023 met inclusion criteria, encompassing a total of 2091 patients from all five continents. The mean graft take rate was 64% in adults and 70% in pediatric patients. Complication events and mortality data are described in the corresponding tables. This systematic review represents the first comprehensive synthesis specifically addressing complications, tumorigenicity, and mortality in burn patients treated with CEAs. The reported adverse events must be interpreted within the context of the high morbidity associated with extensive burn injuries. The observed mortality and number of complications appear primarily related to burn severity rather than to CEA use itself. No adverse events were identified that could be specifically attributed to CEAs. Interpretation of these findings is limited by the clinical and methodological heterogeneity of the included studies. In patients with extensive burn injuries, the use of CEAs is associated with a relatively low number of complications and appears to be a safe and clinically justified treatment option. This study received no internal or external funding. This study is referenced in the PROSPERO Database for systematic reviews under the number CRD420251040971.

Bioengineering doi: 10.3390/bioengineering13101095

Authors: Ruohan Lv Jingwei Feng Yuhui He Yanyong Zhao

Three-dimensional (3D) printing and tissue engineering offer patient-specific strategies for auricular reconstruction, but bibliometric structure alone cannot establish technical maturity or clinical value. We combined bibliometric mapping of English-language Articles and Reviews in the Web of Science Core Collection (2014–2025) with a translational evidence map. The locked corpus comprised 169 publications (143 Articles and 26 Reviews) and 4291 citations at the 12 June 2026 snapshot. Annual output, countries, institutions, authors, journals, highly cited records, author-keyword co-occurrence, and temporal terminology were analyzed. Citation-age sensitivity used citations per publication-year (CPY), year-normalized scores, and within-year percentiles. Record-level coding evaluated technical routes, materials, cell systems, construct scale, evidence stage, scientific endpoints, reporting completeness, and barriers. Among Articles, 40 (28.0%) were in vitro only, 41 (28.7%) reached small-animal testing, four (2.8%) reached large-animal testing, and nine (6.3%) reported human evidence; only one was comparative clinical research. Chondrogenesis or extracellular-matrix outcomes were reported in 84 Articles (58.7%), whereas vascularization, clinical/aesthetic outcomes, and long-term safety/adverse events were each reported in ≤6.3%. The field is expanding internationally but remains predominantly preclinical; progress requires standardized reporting, native-cartilage comparators, long-term shape and safety assessment, and larger-animal and comparative clinical validation.

Bioengineering doi: 10.3390/bioengineering13091094

Authors: Xiaoxue Bai Li Zhao Pan Zhou Ge Bai Longxin Xiao Jianan Chen Ye Yu Weicheng Yan Yang Xu

Core parameters of the Musculoskeletal Ultrasound Computed Tomography System (MUCTS), such as spatial resolution, slice thickness, and three-dimensional reconstruction volume error, directly determine imaging performance and clinical diagnostic accuracy. To objectively and systematically assess the imaging quality and diagnostic reliability of this device, this study took the MUCTS developed by Weishi Medical Imaging Co., Ltd., as the test object. Multidimensional performance tests were carried out using standard ultrasound phantoms and polyurethane-based tissue-mimicking musculoskeletal phantoms. Under unified imaging and reconstruction parameters, key indicators, including effective detection radius, axial resolution, lateral resolution, imaging dead zone, slice thickness, geometric position accuracy, and three-dimensional reconstruction volume error, were quantitatively measured, and compliance verification was performed against product technical specifications. The test results demonstrated that the nominal center frequency was 3 MHz with a measured frequency deviation of 3%, and the effective detection radius reached 90 mm. Within the valid imaging range, the axial resolution was 0.5 mm, lateral resolution was 1 mm, imaging dead zone was 1 mm, and slice thickness was 3.2 mm. The relative error of geometric position accuracy was 3%, the relative deviations in perimeter and area measurements were both −1%, and the relative deviation in three-dimensional reconstructed volume was −8%. All indicators satisfied product technical requirements. Qualitative comparative experiments on tissue-mimicking phantoms confirmed that the system could reproduce multiscale and multi-contrast musculoskeletal anatomical features, and the acquired images showed favorable consistency with designed phantom targets. In conclusion, this musculoskeletal ultrasound tomographic system exhibits stable imaging performance and credible quantitative measurements, and can meet clinical and research demands for tomographic imaging and three-dimensional reconstruction of human musculoskeletal tissues.

Bioengineering doi: 10.3390/bioengineering13091093

Authors: Xiangcheng Lei Ruipeng Li Yangyi Nie Deju Gao Wei Zhang Yuxiao Lai Cairong Li

Magnesium ions (Mg2+) play pivotal roles in bone regeneration by regulating the biological functions of multiple bone repair-related cells. However, the effective concentration ranges and optimal concentration of Mg2+ for different cell types remain incompletely defined, which limits the precise design of Mg-containing biomaterials with controllable ion release for repairing of bone defects. In this study, the concentration-dependent effects of Mg2+ on the biological functions of bone regeneration-related cells were systematically investigated. Human bone marrow mesenchymal stem cells (hBMSCs), human umbilical vein endothelial cells (HUVECs), rat Schwann cells (RSCs), and RAW264.7 macrophage were selected to evaluate the effective concentration of Mg2+ for promoting osteogenesis, angiogenesis, neuroregulation, and immunomodulation, respectively. The results demonstrated that Mg2+ concentrations ranging from to 16 mM, particularly 8 mM, significantly upregulated the expression of M2 macrophage-related markers, thereby promoting the polarization of RAW264.7 macrophages toward the anti-inflammatory M2 phenotype. Moreover, Mg2+ concentrations between 2 and 8 mM effectively enhanced the osteogenic activity of hBMSCs, the angiogenic activity of HUVECs, and the neuroregulatory function of RSCs, as confirmed by immunofluorescence staining and quantitative PCR analysis of osteogenic, angiogenic, and neuroregulation-related marker expression. These findings provide a comprehensive understanding of the concentration-dependent biological effects of Mg2+ on bone regeneration-related cells and establish a theoretical basis for the rational design of next-generation Mg2+-releasing biomaterials with precise biological regulation.

Bioengineering doi: 10.3390/bioengineering13091092

Authors: Zheng Wang Yitong Xia Guifeng Zhang

Machine-learning gait studies often report high accuracy for separating neurodegenerative diseases, but rarely test whether this reflects pathology or the demographic differences that accompany disease in convenience cohorts. We audited the PhysioNet Gait in Neurodegenerative Disease Database (63 subjects: controls, Parkinson’s disease [PD], Huntington’s disease [HD], amyotrophic lateral sclerosis [ALS]), classifying 37 stride-rhythm features under subject-disjoint nested leave-one-subject-out cross-validation. A random forest reached a macro-averaged one-vs-rest AUROC of 0.832 (95% CI 0.759–0.904; permutation p = 0.001). Four complementary audits—demographic-only baselines, confound adjustment, feature-family ablation, and age matching—show that this overstates gait-specific discrimination: five anthropometric variables alone reached an AUROC of 0.812, statistically indistinguishable from the gait model (p = 0.701), and the de-confounded gait model (0.755) fell further to 0.585 once age was matched across groups. HD was the exception, essentially unaffected by adjustment (0.812 → 0.804) and carried by variability and long-range-correlation measures. In an independent, age-matched cohort recorded with a different sensor (n = 110, PD versus control only), the same vulnerability reappeared through a different variable: the gait model (0.828) showed no statistically significant difference from walking speed alone (0.746, p = 0.11), while age and anthropometry fell to chance; this cohort contained no HD subjects, so the HD finding remains externally unvalidated. The shortcut vulnerability generalises across cohorts; the demographic variable that drives it does not. Four-class differential diagnosis is not demonstrated on this benchmark; HD gait arrhythmicity is the one signal that withstands internal sensitivity analysis. We recommend that accuracy claims from small, demographically imbalanced gait cohorts be accompanied by this kind of audit.

Bioengineering doi: 10.3390/bioengineering13091091

Authors: Alexander Fiedler Marius Drysch Sonja Verena Schmidt Pia Weskamp Felix Reinkemeier Flemming Puscz Alexander Sogorski Marcus Lehnhardt Christoph Wallner

Anastomotic angle and flow allocation change together in end-to-side junctions, complicating interpretation of angle-dependent wall shear. We used MITOS Flow Lab, a two-dimensional D2Q9 two-relaxation-time lattice-Boltzmann environment, to examine this coupling in steady, rigid-walled, Newtonian models at Re ≈ 91. Angles of 30–120° were compared under equal outlet pressures and at approximately matched branch-flow fractions of 0.25 and 0.21, achieved with angle-specific static outlet-pressure offsets. Under equal outlet pressures, the branch-flow fraction decreased from 0.356 to 0.151 across this angle range, while the minimum normalized signed recipient-floor shear increased from 0.140 to 0.450. Matching flow allocation substantially reduced angle-associated variation in this endpoint and in the sub-toe response, whereas sub-heel and sub-ostial responses remained angle dependent under the adjusted boundary conditions. This qualitative contrast persisted when the lumen resolution was increased from 32 to 64 nodes for the 0.25 target, although minimum-shear attenuation changed from approximately 83% to 74%. The 0.21 target was examined only on the production grid. These experiments demonstrate that the interpretation of angle-associated shear depends on the flow-allocation condition used for comparison. They do not identify a boundary-independent geometric effect or a causal mediation fraction. Absolute values and attenuation magnitudes remain sensitive to discretization and have not been independently validated. These controlled comparisons provide a framework for interpreting angle-associated shear together with achieved flow allocation and the specified outlet conditions.

Bioengineering doi: 10.3390/bioengineering13091090

Authors: Jeniffer Jesus Pedro Cardoso-Teixeira João Chibante Pedro Dália Meira Ignácio Rodriguez-Uña João Melo Beirão Luís Bastião Silva

Glaucoma requires continuous surveillance, sustained treatment adherence, and timely recognition of postoperative or treatment-related complications. Conventional outpatient follow-up provides structured clinical assessment but offers limited visibility of symptoms between appointments. This study describes TeleGlaukos, a clinician-led adaptive telemonitoring platform designed for hybrid perioperative and therapeutic glaucoma care. Its principal innovation is a configurable protocol engine that transforms static symptom forms into patient-specific workflows through conditional questions, symptom-triggered branching, medication reminders, rule-based alerts, and bidirectional clinician–patient communication. The platform was implemented as a Progressive Web App connected to a modular Django backend through RESTful APIs and WebSocket-enabled communication. The clinical pilot cohort comprised 30 glaucoma patients followed between May 2025 and April 2026. The wider platform environment contained 95 registered accounts: 30 glaucoma patients, 10 ophthalmologists, 10 research or administrative users, 20 development and testing accounts, and 25 non-pilot registered users. Across the platform, 737 engaged sessions, 1329 events, 30 bidirectional messages, and 186 questionnaire responses were recorded. Structured monitoring identified suspected treatment-related adverse reactions in two patients, prompting earlier clinical reassessment and treatment adjustment. Usability was assessed in 22 respondents (18 patients and 4 ophthalmologists), yielding a mean System Usability Scale score of 79.43 ± 17.98 (95% CI: 71.46–87.40), corresponding to a Good adjective rating and an Acceptable usability range; internal consistency was high (Cronbach’s alpha = 0.882). The artificial intelligence-assisted module remained exploratory and was not evaluated as an autonomous clinical component. TeleGlaukos demonstrated technical feasibility, favourable perceived usability, and the capacity of adaptive symptom workflows to surface clinically relevant signals between visits. Larger multicentre studies are required to evaluate effectiveness, safety, adherence, workflow burden, and cost-effectiveness.

Bioengineering doi: 10.3390/bioengineering13091089

Authors: Keith Saum Begoña Campos Diego Celdran Bonafonte Liran Oren Albert Phillip Owens Prabir Roy-Chaudhury

The arteriovenous fistula (AVF) is the preferred method of vascular access for hemodialysis; however, 30–50% of AVFs undergo primary maturation failure and are unsuitable for clinical use. As disturbed hemodynamics initiate endothelial injury and intimal hyperplasia, we designed an endovascular flow-conditioning anastomotic device (FCAD) to directly improve AVF hemodynamics and protect the anastomotic region. Using computational fluid dynamics, we characterized the flow field and wall shear stress (WSS) profiles in idealized AVF models with and without the FCAD. Incorporation of the FCAD into a brachiocephalic AVF model reduced regions of oscillatory WSS and generated a symmetrical flow profile in the draining vein compared to a reference AVF. Parametric studies identified an FCAD geometry with a tab angle, height, and aspect ratio of 30°, 0.1 diameters, and 1.0 that increased time-averaged WSS along the inner venous wall and achieved a physiological WSS (1–3 Pa) without inducing regions of oscillatory flow throughout the cardiac cycle. Similar findings were observed with an in vitro model using particle image velocimetry. This study demonstrates the feasibility of the FCAD to normalize venous flow and WSS while imposing minimal resistance to blood flow. Restoring physiological WSS levels on the venous wall may help preserve endothelial function and improve AVF maturation.

Bioengineering doi: 10.3390/bioengineering13091088

Authors: Sedeek Mosaid Yousif Jihad Nuala Pepper Mohamed Elnemr Ashok Marudanayagam Paul Lee

Fragility fracture and osteoporotic bone loss represent an increasing global burden in ageing populations: the Global Burden of Disease Study 2021 estimated 172.79 million new fractures, 453.31 million prevalent cases, and 25.18 million years lived with disability in 2021, yet the resulting defects are still managed with grafts and scaffolds validated in healthy bone. Osteoporotic bone is not merely rarefied normal bone but a distinct regenerative microenvironment, in which we propose that oxidative stress, cellular senescence, and loss of osteogenesis-coupled type-H vasculature act as a mutually reinforcing triad, superimposed on an adipogenically biassed progenitor pool, blunted osteocyte mechanotransduction and immunosenescence. Additive manufacturing and biofabrication permit patient-specific geometry and spatial and temporal programming of mechanical and biological cues. This review appraises four interdependent design axes—mechanical adaptation through functionally graded, trabecular-mimetic architectures; redox regulation; senescence modulation; and osteoimmunomodulation coupled to vascularisation—distinguishing study type, disease specificity and author-proposed hypotheses. The most direct clinical signal is mechanical: in a retrospective cohort of 163 lumbar interbody fusion levels, printed porous titanium was associated with subsidence in 5.5% of levels versus 24.4% for solid titanium. By contrast, almost all redox, senescence and immunomodulatory scaffold evidence remains preclinical: osteoporosis-specific evidence relies heavily on ovariectomised rodents, and much of the rest comes from non-osteoporotic models. Within the searches described, we identified no completed clinical evaluation with reported results of a multifunctional scaffold in an osteoporosis-defined cohort. As the authors’ proposed conceptual framework rather than an established result, we suggest that these axes be engaged sequentially rather than additively, and that timing, dose and spatial distribution may matter more than the number of bioactivities.

Bioengineering doi: 10.3390/bioengineering13091087

Authors: Fian Dervis Alp Saruhanoglu Duygu Ofluoglu Meltem Koray

Dental injection needle integrity may influence tissue trauma during local anesthesia. This study compared needle-tip microstructural deformation after different local anesthesia techniques using field emission scanning electron microscopy (FESEM). A total of 110 single-use needles were allocated to an unused descriptive control set (n = 5), infiltrative local anesthesia (ILA; n = 35), direct inferior alveolar nerve block anesthesia (DIANBA; n = 35), and indirect inferior alveolar nerve block anesthesia (IIANBA; n = 35). After one clinical application, tips were evaluated for bevel edge irregularity (BEI), tip axis deviation (TAD), and bevel edge microstructural breakdown (BEMB). Inferential analyses were restricted to the three equally sized clinical groups. BEI occurred in 5.7%, 28.6%, and 45.7% of the ILA, DIANBA, and IIANBA groups, respectively (overall p = 0.0007; Cramer’s V = 0.371). After Holm adjustment, BEI differed between ILA and DIANBA (adjusted p = 0.0469) and between ILA and IIANBA (adjusted p = 0.0007). TAD frequencies were 14.3%, 34.3%, and 40.0% (overall p = 0.0466; V = 0.242), but no pairwise comparison remained significant after adjustment. BEMB frequencies were 22.9%, 40.0%, and 45.7%, with no statistically significant overall difference (p = 0.117). Thus, the demonstrated finding is greater needle-tip microstructural deformation, particularly BEI, after nerve block techniques; clinical consequences such as pain or tissue trauma were not assessed and require dedicated investigation.

Bioengineering doi: 10.3390/bioengineering13091086

Authors: Marcus Jäger Andrea Sowislok

Extracellular vesicles (EVs) have emerged as promising cell-free therapeutic agents for bone regeneration due to their ability to modulate osteogenesis, angiogenesis, and immune responses. This review summarizes the current knowledge on EV biology, including their classification, biogenesis, and molecular cargo, with a particular focus on their role in bone healing. EVs derived from mesenchymal stromal cells and other cell types promote osteoblast differentiation, inhibit osteoclast activity, enhance vascularization, and modulate inflammation through the transfer of bioactive proteins and non-coding RNAs. Recent advances in EV engineering, biomaterial-assisted delivery, and isolation techniques are discussed, together with their translational potential for the treatment of critical bone defects and avascular osteonecrosis. Although preclinical evidence is highly encouraging, clinical evidence remains limited, particularly in orthopedics, and clinical translation is challenged by EV heterogeneity, lack of dose standardization, limited understanding of biodistribution, scalable manufacturing, and regulatory requirements. Overall, EV-based therapies represent a promising future strategy for regenerative orthopedics and bone tissue engineering.

Bioengineering doi: 10.3390/bioengineering13091085

Authors: Teodora Ormandzhieva Florian Elger Marcello Silvano Bernd Niemann Eberhard Grambow

(1) Background: Laser in situ fenestration (L-ISF) is an endovascular technique used to preserve target vessel patency in urgent or emergency endovascular aortic repair when fenestrated/branched solutions are unavailable or unsuitable, or as a bailout technique for fenestration or branch misalignment. This on-bench study compared different laser energy levels in five commercially available endografts to assess fabric damage and evaluate fenestration quality, dimensions, and recoil. (2) Methods: A total of 102 fenestrations were created using a 2.0 mm excimer laser catheter at low (45 mJ/mm2) or high (60 mJ/mm2) energy levels and dilated with a 6 mm plain balloon in five polyester endografts (Zenith Alpha, RelayPro, Endurant IIs, Valiant Captivia, Hercules). High-magnification microscopy assessed morphology and fabric damage after puncture, after dilation, and after 24 h. Quantitative measurements included diameters, area, and recoil under standardized study conditions. (3) Results: Valiant exhibited clearer puncture margins and no bulging; Hercules showed frequent tearing and more bulging and warp-aligned recoil at high energy. In Endurant IIs, low-energy L-ISF produced larger diameters and area but also greater recoil. RelayPro revealed stronger surface recoil with high energy. (4) Conclusion: Both low- and high-energy L-ISF were feasible across all tested grafts. A low-energy approach may be preferable for Endurant IIs, potentially reducing the risk of bridging stent graft stenosis.

Bioengineering doi: 10.3390/bioengineering13091084

Authors: Jinglai Lin Linpeng Yao Jianbo Gao Ning Xu Ying Xiong Qi Bai Kang Wang Qi Sun

Accurate preoperative discrimination between indolent and aggressive renal lesions is essential for personalized risk-stratified management and improves the clinical outcomes of renal tumor patients. This study aimed to establish and externally validate a non-invasive computed tomography (CT)-derived radiomic model for identifying aggressive renal tumors and predicting long-term postoperative survival. A total of 3407 patients with pathologically confirmed renal tumors from multiple centers were retrospectively enrolled and divided into three independent cohorts: a training cohort (n = 1840), an internal validation cohort (n = 789), and an external test cohort (n = 778). Based on preoperative multiphasic CT images, a support vector machine (SVM)-based radiomic model was constructed to evaluate tumor aggressiveness. Cox proportional hazards regression analysis was performed to identify independent prognostic factors, and a comprehensive nomogram combining radiomic aggressiveness score and clinicopathological parameters was further developed for survival prediction. The established SVM radiomic model yielded stable discriminatory performance, with an area under the receiver operating characteristic curve (AUC) of 0.740 in the internal validation cohort and 0.763 in the external test cohort. Multivariate Cox regression analysis demonstrated that radiomics-predicted tumor aggressiveness remained an independent adverse prognostic factor for recurrence-free survival (HR = 3.656), overall survival (HR = 1.981), and disease-specific survival (HR = 3.673). Moreover, the integrated nomogram demonstrated favorable calibration for predicting 5-year and 8-year overall survival in the internal validation cohort. This novel CT-based radiomic model provides a robust non-invasive tool for preoperative differentiation of renal tumor aggressiveness. As an independent prognostic biomarker, it can assist clinicians in precise preoperative risk evaluation and facilitate individualized therapeutic decision-making for renal tumor patients.

Bioengineering doi: 10.3390/bioengineering13091083

Authors: Reynalyn Cernechez Seyed Ebrahim Hosseini Muhammad Nadeem Shahbaz Pervez Mohammad Salah Muazma Shahbaz

Depression among students has emerged as a critical issue within educational institutions, leading to the need for approaches that can support early identification of students at risk. This study developed an explainable machine learning framework for predicting student depression risk using non-clinical demographic, academic, lifestyle, and psychosocial factors. Using a public OpenML dataset containing approximately 27,901 student records, logistic regression, random forest, and XGBoost models were developed and evaluated using accuracy, precision, recall, F1-score, and ROC-AUC. Fairness evaluation was conducted across gender and financial stress groups, while SHAP and LIME were applied to provide global, class-level, and local explanations of model predictions. Results showed that all models achieved strong predictive performance, with XGBoost performing best (accuracy = 0.846, recall = 0.883, F1-score = 0.870, ROC-AUC = 0.920). The fairness evaluation showed relatively consistent performance across gender groups, while models performed better in identifying depression cases among students in the high financial stress group. Further, explainability analysis identified Suicidal Thoughts, Academic Pressure, and Financial Stress as the most influential predictors. A Streamlit prototype was developed to demonstrate practical deployment. Overall, the findings demonstrate the potential of explainable machine learning to support student depression risk identification and provide interpretable model predictions.

Bioengineering doi: 10.3390/bioengineering13091082

Authors: Mohammed Alnusayri Sara Mumtaz Bader Aldughayfiq Hisham Allahem Nabil Almashfi Dina Abdulaziz AlHammadi Ahmad Jalal

Cardiovascular disease diagnosis requires accurate and timely analysis of electrocardiogram (ECG) signals to support reliable clinical decision-making. However, ECG signals are inherently non-stationary, exhibit substantial inter-patient variability, and may share similar morphological patterns across different cardiac disorders, making automated multi-class diagnosis challenging. This study proposes a multi-domain machine learning framework for automated ECG-based cardiac disease classification, integrating signal preprocessing, heartbeat segmentation, Variational Mode Decomposition (VMD), multi-domain feature extraction, minimum Redundancy Maximum Relevance (mRMR) feature selection, and hybrid CNN–Transformer learning. Experiments are conducted on the PTB-XL database using five diagnostic superclasses: NORM, MI, STTC, CD, and HYP. First, a fourth-order Butterworth band-pass filter (0.5–40 Hz) is applied to remove baseline wander and high-frequency noise, followed by adaptive Symlet-8 wavelet denoising with soft thresholding to suppress residual high-frequency fluctuations while preserving the P-wave, QRS complex, and T-wave morphology, after which the signal is z-score normalized. R-peaks are subsequently detected to segment standardized cardiac cycles. VMD is then employed to decompose the heartbeat signals into intrinsic modes, from which the most informative modes are retained using correlation-based mode selection. Temporal, statistical, spectral, and nonlinear features are extracted to capture complementary characteristics of cardiac electrical activity, while mRMR selects the eight most informative features by maximizing feature relevance and minimizing redundancy. The resulting representation is processed through a hybrid CNN–Transformer architecture, in which convolutional layers learn local morphological patterns and Transformer-based attention captures long-range dependencies within the cardiac feature representation. The proposed framework achieves 93.60% accuracy, 93.61% macro precision, 93.60% macro recall, 93.60% macro F1-score, and 98.40% macro specificity across the five diagnostic classes. Confusion-matrix analysis, receiver operating characteristic (ROC) analysis, comparative evaluation, and ablation experiments further demonstrate the discriminative capability and robustness of the proposed approach. These findings indicate that multi-domain biomedical feature learning combined with attention-based deep learning can provide an effective and robust strategy for automated ECG-based cardiac disease classification, highlighting the potential of machine learning for intelligent biomedical signal analysis and computer-aided clinical diagnosis.

Bioengineering doi: 10.3390/bioengineering13091081

Authors: Jeong-Hee Seo Eun-Yeong Park Jeong-Hyeon Lee Won-Gi Kim Sung-Jae Lee Bongju Kim

Time-dependent force reduction may affect the biomechanical performance of thermoformed clear aligners. This proof-of-concept in vitro study compared initial and 7-day system-level 3D resultant forces of DURAN, ZENDURA FLX, and MESHEET under attachment-free and attachment-integrated conditions. Thirty independently fabricated aligners (n = 5 per condition) were tested on a rigid resin model with FDI tooth 11 positioned at 5° of labioversion. Forces were recorded at 1 Hz for 168 h under dry room conditions and summarized at 12 h intervals. Temporal series and group-level normalized decay were evaluated descriptively; no longitudinal inferential analysis was performed. Separate 3 × 2 two-way ANOVAs assessed absolute resultant force at 0 and 168 h. Higher endpoint forces were observed under the attachment-integrated condition in this experimental setup at 0 h [F(1,24) = 24.510, p < 0.001, partial η2 = 0.505] and 168 h [F(1,24) = 7.175, p = 0.013, partial η2 = 0.230]. However, because testing followed a fixed, nonrandomized sequence, the observed difference cannot be attributed exclusively to attachment condition. Material and interaction effects were not significant at either endpoint. Under the attachment-integrated condition, normalized decay was 51.4%, 48.2%, and 44.5%, while 168 h mean forces were 1.415, 1.664, and 1.558 N, respectively. These findings do not establish decay-rate differences, material superiority, equivalence, or clinical efficacy.

Bioengineering doi: 10.3390/bioengineering13091080

Authors: Xinyue Liu Huijuan Shi Xi Gong Huimeng Chen Yujiao Wang Xiaohan Zhao Shuang Ren Hui Liu

Individuals undergoing anterior cruciate ligament reconstruction (ACLR) remain at risk of secondary injury and joint degeneration, highlighting the need for functional assessments that detect abnormal load regulation and compensatory movement strategies. This study compared lower-limb kinematics, kinetics, and joint work contributions during double-leg squatting between 27 patients 5.9 ± 0.7 months after ACLR and 23 healthy controls. Three-dimensional marker trajectories and ground reaction forces were collected, and biomechanical variables were analyzed using a two-way mixed-design ANOVA. The reconstructed limb showed smaller knee flexion angles and lower knee extensor moments than both the contralateral limb and the non-dominant limb of healthy controls (p ≤ 0.015), together with reduced ankle dorsiflexion and peak vertical ground reaction force versus the contralateral limb (p < 0.001). Compared with healthy controls, the ACLR group also showed altered frontal- and transverse-plane hip and knee mechanics (p ≤ 0.017). The reconstructed limb demonstrated greater negative hip work and hip work contribution, lower negative knee work and knee work contribution (all p ≤ 0.003), and greater ankle work contribution than the contralateral limb (p = 0.018). These findings describe a cross-sectional biomechanical pattern at approximately 6 months after ACLR characterized by sagittal-plane knee control deficits, bilateral hip control differences, and altered inter-joint work distribution.

Bioengineering doi: 10.3390/bioengineering13091079

Authors: Xinyu Zhang Jianwei Cui Yuxiang Dai Wenyi Zhang

Ultrasound assessment of developmental dysplasia of the hip (DDH) in infants is highly operator-dependent, particularly during standard plane acquisition and Graf angle measurement. This paper proposes DDH-Net, a fully automated computer-aided assessment method for infant hip ultrasound that follows the complete Graf workflow. First, YOLOv8n-pose is used to detect eight anatomical structures and two ilium orientation keypoints for standard plane determination. Regions of interest (ROIs) are then cropped according to the detection results, and U-Net is employed to perform fine segmentation of the ilium, labrum, and lower limb of the ilium. Finally, Graf reference lines are constructed from the segmentation results to enable automatic measurement of the α and β angles. The study included 1409 infants, comprising 2648 ultrasound images and 50 ultrasound videos. On an independent test set of 529 images, the accuracy, sensitivity, and specificity of standard plane detection were 96.8%, 94.5%, and 100.0%, respectively. The model automatically saved 187 candidate frames from the videos, of which 93.1% were rated by experts as having no or only minor clinically relevant discrepancies; 48 of the 50 videos (96.0%) contained at least one clinically acceptable candidate frame. Compared with full-image segmentation, ROI-based segmentation reduced processing time by 33.7%. In 307 standard plane images, the mean absolute errors (MAE) for the α and β angles were 1.61° and 2.13°, respectively, with corresponding intraclass correlation coefficients (ICC) of 0.913 and 0.766. The complete pipeline achieved a processing speed of 28.87 frames per second, indicating that DDH-Net has the potential to provide efficient and interpretable analysis of infant hip ultrasound images.

Bioengineering doi: 10.3390/bioengineering13091078

Authors: Xiaonan Gong Jiang Heng Hao Feng Suzhan Zhang

Breast-conserving surgery (BCS) requires accurate intraoperative localization of the tumor, which moves between the prone diagnostic and supine surgical positions. We present a projector-based augmented-reality (AR) localization system that projects a crosshair directly onto the patient’s skin at the predicted tumor location. The pipeline combines rigid registration of ceramic fiducials between prone and supine CT (tumor error 0.30 mm on a real phantom), a SOFA finite-element simulation of gravity-driven prone-to-supine soft-tissue deformation (mean bead displacement 24.7 mm on a synthetic mesh; not validated on a deformable phantom), two-stage ICP surface alignment (RMSE 0.92 mm on a simulated camera; not yet validated on real D415 acquisitions), and projector–camera (ProCam) extrinsic calibration driving the crosshair projection. Because the tumor and sub-surface fiducials are not visible in the live RGB-D scene, a direct non-circular in vivo target registration error (TRE) on the tumor cannot be measured in the present phantom configuration. We instead report a component-based end-to-end error budget obtained as the root-sum-square of three independently measured component errors: 2.7 mm (RSS planning estimate, not a directly measured end-to-end accuracy) —rigid CT 0.30 mm, ProCam physical projection 2.50 mm, ICP 0.92 mm. The ProCam term is largest but sub-3 mm. The Gray-code reprojection residual (~11 mm in 15-pose recalibration; 8.62 mm in the deployed 4-pose calibration) overestimates the actual projection error by ~4× (resp. ~3.4×): least-squares calibration averages the consumer-DLP vertical quantization (16-level Gray-code decode on XPR-DLP) into an accurate rigid transform, so the per-observation residual reflects observation quantization, not projection accuracy. The quantization’s measurable cost is projection variance (max 5.5 mm), which motivates phase-shifting structured light. Boundary conditions on phantom rigidity, sub-surface fiducials, and simulated-camera ICP are stated explicitly.

Bioengineering doi: 10.3390/bioengineering13091077

Authors: John A. Dangerfield Arkar W. Maung Poovasit Klinoubol

Background: Chronic wounds that fail to heal despite debridement and advanced care are an unmet clinical need, particularly in elderly patients and those with peripheral arterial disease, for whom the alternatives are often indefinite maintenance or amputation. The mesenchymal stromal cell (MSC) secretome is pro-reparative in preclinical models, but human evidence is concentrated on shorter-duration diabetic foot ulcers and has usually reported area reduction rather than complete closure. Methods: We describe two patients with long-standing, treatment-refractory wounds treated with a cell-free Wharton’s jelly-derived MSC secretome delivered in a thermoresponsive polymeric carrier that gels at the wound bed, applied every three days alongside standard care. Results: The cases comprise the following: Case 1, a 95-year-old woman with peripheral arterial disease and a two-year ankle ulcer refractory to multiple advanced products; and Case 2, a 69-year-old man with a one-year neuro-ischemic diabetic foot ulcer persisting after amputation and repeated debridement. Both achieved complete closure at approximately 27 and 28 weeks, respectively, from baseline wound areas of approximately 7.5 cm2 (Case 1) and 12 cm2 (Case 2), using approximately 20 mL of preparation each. Closure was sustained at the latest follow-up (27 March 2026, Case 1; 21 June 2026, Case 2). Treatment was well tolerated; Case 2 had an intercurrent infection requiring a brief pause. Conclusions: A two-patient series cannot establish efficacy, but complete closure in wounds this refractory supports controlled evaluation of this cell-free platform, with complete closure as the primary endpoint.

Bioengineering doi: 10.3390/bioengineering13091076

Authors: Eun-Jin Lee Longyi Chen Zachary Krassin Sabrina Herrmann Gretchen J. Mahler Mandy B. Esch

The interaction of chemicals, nanoparticles, and circulating cells with the endothelium depends on the magnitude of the mechanical shear produced by the flow of blood. When simulating those interactions with microphysiological systems (MPSs), it is critical to reproduce those shear conditions faithfully. For example, unidirectional flow of a specific magnitude keeps the endothelium healthy with normal barrier tissue function, while bidirectional flow mimics disease conditions with compromised barrier function. Additionally, in MPS, recirculating fluid may be necessary to retain tissue-derived factors and metabolites. However, existing MPS designs struggle to achieve medium recirculation of small volumes of liquid with precise flow control. Here, we present an MPS design that is highly versatile and overcomes this limitation. We demonstrate how the device can produce a wide range of fluidic flow rates that can accommodate both low shear conditions suitable for tissues that typically are only exposed to interstitial flow and high shear conditions suitable for barrier tissues that experience blood flow. We demonstrate the device’s functionality by culturing human umbilical vein endothelial cells (HUVEC) and confirming their flow-aligned morphology through immunostaining of the adherens junction protein (VE-cadherin) and actin filaments. Furthermore, we present a mathematical model that can be used to calculate operating parameters for culturing any tissue under optimum conditions. We also discuss how the device can be adjusted to recirculate liquid volumes ranging from 100 µL to 5 mL. This versatile system holds promise for commercial applications, including the investigation of expensive compounds that are limited to very small volume samples such as rare cells (e.g., circulating tumor cells) or engineered therapeutic cells with barrier tissues. By offering precise control over a wide range of flow conditions with medium recirculation of small liquid volumes, our device addresses a critical gap in current MPS technology.