Search Results (2,928)

This review explores how multimodal foundation models (MFMs) are transforming biliary tract cancer (BTC) research. BTCs are aggressive malignancies with poor prognosis, presenting unique challenges due to difficult diagnostic methods, molecular complexity, and rarity. Importantly, intrahepatic cholangiocarcinoma (iCCA), perihilar cholangiocarcinoma (pCCA), and distal bile duct cholangiocarcinoma (dCCA) represent fundamentally distinct clinical entities, with iCCA presenting as mass-forming lesions amenable to biopsy and targeted therapies, while pCCA manifests as infiltrative bile duct lesions with challenging diagnosis and primarily palliative management approaches. MFMs offer potential to advance research by integrating radiological images, histopathology, multi-omics profiles, and clinical data into unified computational frameworks, with applications tailored to these distinct BTC subtypes. Key applications include enhanced biomarker discovery that identifies previously unrecognizable cross-modal patterns, potential for improving currently limited diagnostic accuracy—though validation in BTC-specific cohorts remains essential—accelerated drug repurposing, and advanced patient stratification for personalized treatment. Despite promising results, challenges such as data scarcity, high computational demands, and clinical workflow integration remain to be addressed. Future research should focus on standardized data protocols, architectural innovations, and prospective validation studies. The integration of artificial intelligence (AI)-based methodologies offers new solutions for these historically challenging malignancies. However, current evidence for BTC-specific applications remains largely theoretical, with most studies limited to proof-of-concept designs or related cancer types. Comprehensive clinical validation studies and prospective trials demonstrating patient benefit are essential prerequisites for clinical implementation. The timeline for evidence-based clinical adoption likely extends 7–10 years, contingent on successful completion of validation studies addressing current evidence gaps. Full article

This study presents a novel deep learning framework for classifying visual images based on brain responses recorded through electroencephalogram (EEG) signals. The primary challenge in EEG-based visual pattern recognition lies in the inherent spatiotemporal variability of neural signals across different individuals and recording sessions, which severely limits the generalization capabilities of classification models. Our work specifically addresses the task of identifying which image category a person is viewing based solely on their recorded brain activity. The proposed methodology incorporates three primary components: first, a brain hemisphere asymmetry-based dimensional reduction approach to extract discriminative lateralization features while addressing high-dimensional data constraints; second, an advanced channel selection algorithm utilizing Fisher score methodology to identify electrodes with optimal spatial representativeness across participants; and third, a Dynamic Temporal Warping (DTW) alignment technique to synchronize temporal signal variations with respect to selected reference channels. Comprehensive experimental validation on a visual image classification task using a publicly available EEG-based visual classification dataset, ImageNet-EEG, demonstrates that the proposed disambiguation framework substantially improves classification accuracy while simultaneously enhancing model convergence characteristics. The integrated approach not only outperforms individual component implementations but also accelerates the learning process, thereby reducing training data requirements for EEG-based applications. These findings suggest that systematic spatiotemporal disambiguation represents a promising direction for developing robust and generalizable EEG classification systems across diverse neurological and brain–computer interface applications. Full article

Finite element (FE) modeling has emerged as a powerful computational approach in cardiovascular biomechanics, enabling detailed simulations of myocardial stress, strain, and hemodynamics, which are challenging to measure with conventional imaging techniques. This narrative review explores the progression of cardiac FE modeling from research-focused applications to its increasing integration into clinical practice. Specific attention is given to the mechanical effects of myocardial infarction, the limitations of conventional LV volume-reduction surgeries, and novel therapeutic approaches like passive myocardial reinforcement via hydrogel injections. Furthermore, the review highlights the critical role of patient-specific FE simulations in optimizing LV assist device parameters and guiding targeted device placements. Cutting-edge developments in artificial intelligence-enhanced FE modeling, including surrogate models and precomputed simulation databases, are examined for their potential to facilitate real-time, personalized therapeutic decision-making. Collectively, these advancements position FE modeling as an essential tool in precision medicine for structural heart disease. Full article

Cardiovascular disease remains the world’s leading cause of mortality, yet everyday care still relies on episodic, symptom-driven interventions that detect ischemia, arrhythmias, and remodeling only after tissue damage has begun, limiting the effectiveness of therapy. A narrative review synthesized 183 studies published between 2016 and 2025 that were located through PubMed, MDPI, Scopus, IEEE Xplore, and Web of Science. This review examines CVD diagnostics using innovative technologies such as digital cardiovascular twins, which involve the collection of data from wearable IoT devices (electrocardiography (ECG), photoplethysmography (PPG), and mechanocardiography), clinical records, laboratory biomarkers, and genetic markers, as well as their integration with artificial intelligence (AI), including machine learning and deep learning, graph and transformer networks for interpreting multi-dimensional data streams and creating prognostic models, as well as generative AI, medical large language models (LLMs), and autonomous agents for decision support, personalized alerts, and treatment scenario modeling, and with cloud and edge computing for data processing. This multi-layered architecture enables the detection of silent pathologies long before clinical manifestations, transforming continuous observations into actionable recommendations and shifting cardiology from reactive treatment to predictive and preventive care. Evidence converges on four layers: sensors streaming multimodal clinical and environmental data; hybrid analytics that integrate hemodynamic models with deep-, graph- and transformer learning while Bayesian and Kalman filters manage uncertainty; decision support delivered by domain-tuned medical LLMs and autonomous agents; and prospective simulations that trial pacing or pharmacotherapy before bedside use, closing the prediction-intervention loop. This stack flags silent pathology weeks in advance and steers proactive personalized prevention. It also lays the groundwork for software-as-a-medical-device ecosystems and new regulatory guidance for trustworthy AI-enabled cardiovascular care. Full article

Artificial intelligence (AI) is now the computational core of smart building automation, acting across the entire cyber–physical stack. This review surveys peer-reviewed work on the integration of AI with indoor environmental quality (IEQ) and energy performance, distinguishing itself by presenting a holistic synthesis of the complete technological evolution from IoT sensors to generative AI. We uniquely frame this progression within a human-centric architecture that integrates digital twins of both the building (DT-B) and its occupants (DT-H), providing a forward-looking perspective on occupant comfort and energy management. We find that deep reinforcement learning (DRL) agents, often developed within physics-calibrated digital twins, reduce annual HVAC demand by 10–35% while maintaining an operative temperature within ±0.5 °C and CO₂ below 800 ppm. These comfort and IAQ targets are consistent with ASHRAE Standard 55 (thermal environmental conditions) and ASHRAE Standard 62.1 (ventilation for acceptable indoor air quality); keeping the operative temperature within ±0.5 °C of the setpoint and indoor CO₂ near or below ~800 ppm reflects commonly adopted control tolerances and per-person outdoor air supply objectives. Regarding energy impacts, simulation studies commonly report higher double-digit reductions, whereas real building deployments typically achieve single- to low-double-digit savings; we therefore report simulation and field results separately. Supervised learners, including gradient boosting and various neural networks, achieve 87–97% accuracy for short-term load, comfort, and fault forecasting. Furthermore, unsupervised models successfully mine large-scale telemetry for anomalies and occupancy patterns, enabling adaptive ventilation that can cut sick building complaints by 40%. Despite these gains, deployment is hindered by fragmented datasets, interoperability issues between legacy BAS and modern IoT devices, and the computer energy and privacy–security costs of large models. The key research priorities include (1) open, high-fidelity IEQ benchmarks; (2) energy-aware, on-device learning architectures; (3) privacy-preserving federated frameworks; (4) hybrid, physics-informed models to win operator trust. Addressing these challenges is pivotal for scaling AI from isolated pilots to trustworthy, human-centric building ecosystems. Full article

Background/Objetives: Idiopathic pulmonary fibrosis (IPF) is a progressive interstitial lung disease with poor prognosis. Nutritional disorders, particularly cachexia, significantly impact morbidity and mortality in IPF but remain under-investigated. This study aimed to characterize cachexia phenotypes in IPF through morphofunctional assessment (MFA) and to evaluate their prognostic relevance, including the role of mitochondrial biomarkers. Methods: In this prospective bicenter study, 85 IPF patients underwent MFA including bioelectrical impedance vector analysis (BIVA), nutritional ultrasound (NU), and T12-level computed tomography (T12-CT) for body composition. Functional and strength assessments included timed up and go test (TUG) and handgrip strength (HGS), respectively. Cachexia was defined by Evans’ criteria, Martin’s CT-based criteria, and our IPF-specific proposed definition. Serum GDF-15 and PGC-1α levels were also measured. Results: Cachexia prevalence varied by definition: 24.71% (Evans), 29.5% (Martin) and 42.4% (IPF Cachexia Syndrome). Cachectic patients showed significantly lower muscle mass, function, and quality (measured by reduced muscle attenuation at T12-CT), along with higher GDF-15 and lower PGC-1α levels. The presence of IPF Cachexia syndrome (HR 2.56; 95% CI, 1.08–6.07; p = 0.033), GDF-15 > 4412.0 pg/mL (HR 3.21; 95% CI, 1.04–9.90; p = 0.042) and impaired TUG (>8 s) (HR 3.77; 95% CI, 1.63–8.71; 0.002) were all independently associated with increased 24-month mortality. Conclusions: Cachexia is prevalent in IPF and showed strong concordance between the three diagnostic criteria. The IPF Cachexia syndrome, based on comprehensive morphofunctional phenotyping, demonstrated superior discriminatory capacity. The addition of mitochondrial biomarkers may improve early detection and support personalized interventions to improve patient outcomes. Full article

Nutrition plays a pivotal role in preventive health, yet existing digital solutions often lack personalization and accessibility. This study presents an AI-driven framework that integrates machine learning (ML) and natural language processing (NLP) to deliver dynamic, user-centric dietary recommendations. A gradient boosting model, trained on NHANES demographic and anthropometric data, predicts caloric needs with an MAE of 132 kcal, while a locally deployed LLM (Mistral 7B) interprets free-text dietary constraints with 91% accuracy. Rule-based filtering from the USDA database ensures nutritional balance. A pilot usability test (n = 5) confirmed the system’s practicality and satisfaction. The proposed framework addresses key gaps in scalability, privacy, and adaptability, demonstrating the potential of hybrid AI techniques in applied nutrition science. By bridging computational methods with food science, this work offers a reproducible, modular solution for personalized health applications. Full article

Background: Multiple Myeloma (MM) is a malignant plasma cell dyscrasia that progresses through the consecutive asymptomatic, often undiagnosed, precancerous stages of Monoclonal Gammopathy of Undetermined Significance (MGUS) and Asymptomatic Multiple Myeloma (SMM). MM is characterized by low survival rates, severe complications and drug resistance; therefore, understanding the molecular mechanisms of progression is crucial. This study aims to detect genetic mutations, both germline and somatic, that contribute to disease progression and drive tumorigenesis at the final stage of MM, using samples from patients presenting MGUS or SMM, and newly diagnosed MM patients. Methods: Mutations were identified through a fully computational pipeline, implemented in a Linux and RStudio environment, applied to each patient sequence, obtained through single-cell RNA-sequencing (scRNA-seq), separately. Structural and functional mutation types were identified by stage, along with the affected genes. The analysis included quality control, removal of the Unique Molecular Identifiers (UMIs), trimming, genome mapping and result visualization. Results: The findings revealed frequent germline and somatic mutations, with distinct structural and functional patterns across disease stages. Mutations in key genes were identified, pointing to molecules that may play a central role in carcinogenesis and disease progression. Notable examples include the HLA-A, HLA-B and HLA-C genes, as well as the KIF, EP400 and KDM gene families, with the first four already confirmed. Comparative analysis between the stages highlighted molecular transition events from one stage to another. Emphasis was given to novel genes discovered in newly diagnosed MM patients, that might contribute to the tumorigenesis that takes place. Conclusions: This study contributes to the understanding of the genetic basis of plasma cell dyscrasias and the transition events between the stages, offering insights that could aid in early detection and diagnosis, guide the development of personalized therapeutic strategies, and improve the understanding of mechanisms responsible for resistance to existing therapies. Full article

Human keypoint detection has become a fundamental task in computer vision, underpinning a wide range of downstream applications such as action recognition, intelligent surveillance, and human–computer interaction. Accurate localization of keypoints is crucial for understanding human posture, behavior, and interactions in various environments. In this paper, we propose a deep-learning-based human skeletal keypoint detection framework that leverages a High-Resolution Network (HRNet) to achieve robust and precise keypoint localization. Our method maintains high-resolution representations throughout the entire network, enabling effective multi-scale feature fusion, without sacrificing spatial details. This approach preserves the fine-grained spatial information that is often lost in conventional downsampling-based methods. To evaluate its performance, we conducted extensive experiments on the COCO dataset, where our approach achieved competitive performance in terms of Average Precision (AP) and Average Recall (AR), outperforming several state-of-the-art methods. Furthermore, we extended our pipeline to support multi-person keypoint detection in real-time scenarios, ensuring scalability for complex environments. Experimental results demonstrated the effectiveness of our method in both single-person and multi-person settings, providing a comprehensive and flexible solution for various pose estimation tasks in dynamic real-world applications. Full article

Facial features are critical visual indicators for understanding what a person is experiencing, providing valuable insights into their emotions and physical states. However, accurately detecting these features under diverse conditions remains a significant challenge, especially in computationally constrained environments. This paper presents a facial feature extraction method designed to identify regions of interest for detecting facial cues, with a focus on improving the accuracy of eye and mouth detection. Addressing the limitations of standard Haar cascade classifiers, particularly in challenging scenarios such as droopy eyes, red eyes, and droopy mouths, this method introduces a correction algorithm rooted in normal human facial anatomy, emphasizing symmetry and consistent feature placement. By integrating this correction algorithm with a feature-based refinement process, the proposed approach enhances detection accuracy from 67.22% to 96.11%. Through this method, the accurate detection of facial features like the eyes and mouth is significantly improved, offering a lightweight and efficient solution for real-time applications while maintaining computational efficiency. Full article

Background: By leveraging advanced wireless technologies, Healthcare Industry 5.0 promotes the continuous monitoring of real-time medical acquisition from the physical environment. These systems help identify early diseases by collecting health records from patients’ bodies promptly using biosensors. The dynamic nature of medical devices not only enhances the data analysis in medical services and the prediction of chronic diseases, but also improves remote diagnostics with the latency-aware healthcare system. However, due to scalability and reliability limitations in data processing, most existing healthcare systems pose research challenges in the timely detection of personalized diseases, leading to inconsistent diagnoses, particularly when continuous monitoring is crucial. Methods: This work propose an adaptive and secure framework for disability identification using the Internet of Medical Things (IoMT), integrating edge computing and artificial intelligence. To achieve the shortest response time for medical decisions, the proposed framework explores lightweight edge computing processes that collect physiological and behavioral data using biosensors. Furthermore, it offers a trusted mechanism using decentralized strategies to protect big data analytics from malicious activities and increase authentic access to sensitive medical data. Lastly, it provides personalized healthcare interventions while monitoring healthcare applications using realistic health records, thereby enhancing the system’s ability to identify diseases associated with chronic conditions. Results: The proposed framework is tested using simulations, and the results indicate the high accuracy of the healthcare system in detecting disabilities at the edges, while enhancing the prompt response of the cloud server and guaranteeing the security of medical data through lightweight encryption methods and federated learning techniques. Conclusions: The proposed framework offers a secure and efficient solution for identifying disabilities in healthcare systems by leveraging IoMT, edge computing, and AI. It addresses critical challenges in real-time disease monitoring, enhancing diagnostic accuracy and ensuring the protection of sensitive medical data. Full article

Litz wires are extensively employed in contemporary high-frequency switching power electronics to mitigate conductor losses. Minimizing additional winding losses caused by high-frequency phenomena, such as skin and proximity effects, is a critical design consideration for achieving high power density in modern power electronics. However, accurately predicting losses in structures composed of numerous twisted and insulated strands remains a challenge. With the increasing accessibility of commercial numerical tools, such as finite element method (FEM) solvers, simulation-based approaches have become indispensable tools for analyzing electromagnetic phenomena in complex magnetic device structures under high-frequency conditions. In parallel, data-driven modeling has emerged as a powerful method, enabling pattern identification based on datasets; however, such approaches rely on the availability of large amounts of reliable high-quality data. Generating such large-scale FEM datasets, however, is often constrained by long computation times and high memory consumption. Despite the remarkable advancements in computing power, full three-dimensional (3D) FEM analysis at the strand level for Litz wire windings often remains infeasible within personal computing environments. To address these challenges, this study presents a computationally efficient two-dimensional FEM-based framework that integrates a data-driven fitting model with optimized geometric discretization and meshing strategies, enabling accurate analysis with reduced computational load. The proposed approach, which incorporates optimal meshing conditions into commercially available 2D FEM tools and a simple data-driven fitting model, enables accurate prediction of the frequency-dependent AC resistance of multi-turn Litz windings using a typical personal computer. Its feasibility is further demonstrated through experimental frequency response measurements on both 12-turn and 21-turn windings fabricated with 150-strand Litz wire, which show strong agreement with the corrected simulation results, confirming the model’s accuracy and practical applicability. Full article

Virtual Reality (VR) is a useful educational tool in healthcare, allowing students to practise and improve practical skills. In speech therapy (ST), the need to revise academic curricula to adapt them to university contexts and integrate them into advanced clinical practices has highlighted the need to analyse the use of VR in this sector. The objective of this scoping review was to investigate whether research has considered using VR to support ST students’ training and highlight potential gaps in the literature. The study followed the JBI methodology for scoping reviews and was reported according to PRISMA-ScR guidelines. A protocol to conduct the current review was developed and registered on the Open Science Framework. The articles considered were retrieved from databases specialising in healthcare, computer science, and education, and were enhanced by results found with the help of AI-based tools. No constraints were applied and all study types were considered. Fourteen studies were included in the review and analysed under four core subjects: VR technology, ST context, training purposes, and main outcomes and assessment methods. The VR types identified in the studies were grouped into four categories, i.e., non-immersive VR (6/14, 42.9%), immersive VR (5/14, 35.7%), non-specified VR type (2/14, 14.3%), and semi-immersive VR (1/14, 7.1%). Most studies (5/14, 35.7%) focused on clinical skills acquisition, others addressed communication and interpersonal collaborative skills (3/14, 21.4%), while the remaining focused on person-centred care and awareness, clinical interviewing or reasoning skills, and performance knowledge (2/14 each, 14.3%). VR is still in its early stages in ST education. Some recent studies suggest VR supports students’ communication, interdisciplinary, and clinical skills. Although still limited in the context of ST education, the increasing affordability and ease of development of VR, along with its growing use in other healthcare fields, suggest that its underuse might be due to institutional barriers and lack of standardised frameworks. Overall, the findings suggest that VR offers promising support for experiential and skills-based learning. Full article

The order picking process in Goods-to-Person (G2P) systems involves a set of interdependent yet often separately addressed decisions, such as order allocation, sequencing, and rack handling. This study focuses on the joint optimization of order allocation, order sequencing, rack selection, and rack sequencing in a G2P robotic mobile fulfillment system with multiple picking stations. To model this complex problem, we develop a mathematical formulation and propose a two-phase heuristic algorithm that combines simulated annealing, genetic algorithms, and beam search for efficient solution. In addition, we explore and compare two order allocation strategies—order similarity and order association—across a range of operational scenarios. Extensive computational experiments and sensitivity analyses demonstrate the effectiveness of the proposed approach and provide insights into how strategic order allocation can significantly improve picking efficiency. Computational experiments on small-scale instances show that our algorithm achieves near-optimal solutions with up to 93.3% reduction in computation time compared to exact optimization for small cases. In large-scale scenarios, the order similarity strategy reduces rack movements by up to 44.8% and the order association strategy by up to 33.5% relative to a first-come, first-served baseline. Sensitivity analysis reveals that the association strategy performs best with fewer picking stations and lower rack capacity, whereas the similarity strategy is superior in systems with more stations or higher rack capacity. The findings offer practical guidance for the design and operation of intelligent warehousing systems. Full article

(1) Background: Response to neoadjuvant chemotherapy (NAC) is a key prognostic indicator in breast cancer. However, current response evaluation methods rely on histopathological assessment after surgery, delaying opportunities for early treatment adaptation. This study aimed to develop a machine learning model by integrating radiomic features extracted from pre-treatment, contrast-enhanced computed tomography (CT) images with baseline clinical variables to predict NAC response before therapy initiation. (2) Methods: The study investigated two categories of response: (i) pathologic complete response (pCR) versus non-pCR, and (ii) clinical response versus non-response, where clinical response was defined as a reduction in tumor size of at least 30%, encompassing both complete and partial responses. Radiomic features (n = 214) were extracted from intratumoral and peritumoral regions of pre-treatment CT images. Clinical variables (n = 7) were also incorporated to enhance predictive capability. A predictive model was developed using XGBoost algorithm, and performance was evaluated across ten independent data partitions using metrics including accuracy, precision, sensitivity, specificity, F1-score, and AUC. (3) Results: A total of 177 patients were enrolled in the study. The combined clinical-radiomic model set exhibited superior predictive performance compared to models based solely on either radiomic or clinical features. For pCR classification, integrating clinical and radiomic features produced the strongest model, achieving 82.8% accuracy with an AUC of 0.846. The clinical model alone reached 71.4% accuracy and an AUC of 0.797, while the radiomic model achieved 67.5% accuracy and an AUC of 0.615. For clinical response classification, the combined model again outperformed the individual models, achieving 71.7% accuracy with an AUC of 0.725, compared with 65.0% accuracy and an AUC of 0.666 for the clinical model, and 65.6% accuracy with an AUC of 0.615 for the radiomic model. (4) Conclusions: These results demonstrate that integrating CT radiomic features with clinical information enhances the prediction of NAC response, supporting the potential for earlier and more personalized therapeutic decision-making in breast cancer management. Full article