Search Results (1,902)

Ovarian cancer’s high mortality is primarily due to late-stage diagnosis, underscoring the critical need for improved early detection tools. This study develops and validates explainable artificial intelligence (XAI) models to discriminate malignant from benign ovarian masses using readily available demographic and laboratory data. A dataset of 309 patients (140 malignant, 169 benign) with 47 clinical parameters was analyzed. The Boruta algorithm selected 19 significant features, including tumor markers (CA125, HE4, CEA, CA19-9, AFP), hematological indices, liver function tests, and electrolytes. Five ensemble machine learning algorithms were optimized and evaluated using repeated stratified 5-fold cross-validation. The Gradient Boosting model achieved the highest performance with 88.99% (±3.2%) accuracy, 0.934 AUC-ROC, and 0.782 Matthews correlation coefficient. SHAP analysis identified HE4, CEA, globulin, CA125, and age as the most globally important features. Unlike black-box approaches, our XAI framework provides clinically interpretable decision pathways through LIME and SHAP visualizations, revealing how feature values push predictions toward malignancy or benignity. Partial dependence plots illustrated non-linear risk relationships, such as a sharp increase in malignancy probability with CA125 > 35 U/mL. This explainable approach demonstrates that ensemble models can achieve high diagnostic accuracy using routine lab data alone, performing comparably to established clinical indices while ensuring transparency and clinical plausibility. The integration of state-of-the-art XAI techniques highlights established biomarkers and reveals potential novel contributors like inflammatory and hepatic indices, offering a pragmatic, scalable triage tool to augment existing diagnostic pathways, particularly in resource-constrained settings. Full article

The paper presents designing thermomechanical processing routes for medium-carbon boron-bearing microalloyed steel and investigates their effect on microstructure–property characteristics obtained through controlled cooling directly from hot forging temperature. Direct cooling was carried out in situ within the industrial process of hot forging, replacing conventional heat treatment with slow and accelerated air cooling, realized with a fully automated fan-cooling laboratory conveyor which accommodates the desired cooling strategy. Comparative analysis of conventionally normalized and direct-cooled microstructure and mechanical properties obtained under varied thermo-mechanical conditions is presented to investigate the potential of medium-carbon microalloyed steel with boron addition for producing tailored properties comparable to those of the normalized condition. The obtained microstructure composed of grain-boundary ferrite and pearlite which resulted in tensile properties as good as Re ≈ 610 MPa, Rm ≈ 910 MPa, and elongation A5 ≥ 12%. Although the achieved microstructure–property parameters differ from those achieved through conventional normalizing (Rm ≤ 780 MPa, Re ≤ 460 MPa, and A ≥ 14%), they are considerable in terms of selected machinability aspects. The observed effect of the imposed treatment strategies on interlamellar spacing and morphology of ferrite showed possibilities regarding the control of mechanical properties and application of direct cooling as a beneficial alternative to conventional normalizing, where energy consumption is the main concern in manufacturing high-duty parts made of boron-bearing microalloyed steel 35MnTiB4. Full article

Three-dimensional fragment assembly technology has significant application value in fields such as cultural relic restoration, medical image analysis, and industrial quality inspection. To address the common challenges of limited feature representation ability and insufficient assembling accuracy in existing methods, this paper proposes a geometry-aware hierarchical fragment assembly framework (GeoAssemble). The core contributions of our work are threefold: first, the framework utilizes DGCNN to extract local geometric features while integrating centroid relative positions to construct a multi-dimensional feature representation, thereby enhancing the identification quality of fracture points; secondly, it designs a two-stage matching strategy that combines global shape similarity coarse matching with local geometric affinity fine matching to effectively reduce matching ambiguity; finally, we propose an auxiliary transformation estimation mechanism based on the geometric center of fracture point clouds to robustly initialize pose parameters, thereby improving both alignment accuracy and convergence stability. Experiments conducted on both synthetic and real-world fragment datasets demonstrate that this method significantly outperforms baseline methods in matching accuracy and exhibits higher robustness in multi-fragment scenarios. Full article

A modular, multi-functional (encompassing data acquisition, management, preprocessing, and transmission) sensing (MMFS) system based upon the Internet of Things (IoT) paradigm is discussed in this paper with the goal of continuous real-time, multi-sensor and multi-location monitoring of aircraft (including drones) structural performances during flight. According to industrial and system requirements, a microcontroller and four sensors (strain, acceleration, vibration, and temperature) were selected and integrated into the system. To enable the determination of potential in-flight failures and estimates of the remaining useful service life of the aircraft, resistance strain gauge networks, piezoelectric sensors for capturing structural vibrations and impact, accelerometers, and thermistors have been integrated into the MMFS system. Real flight tests with Evektor’s Cobra VUT100i and SportStar RTC aircraft have been undertaken to demonstrate the features of recorded data and provide requirements for the MMFS functional design. Real flight test data were analysed, indicating that a sampling rate of 1000 Hz is necessary to balance representation of relevant features within the data and potential loss of quality in fatigue life estimation. The design and evaluation of the performance of a prototype (evaluated via representative stress/strain experiments using an Instron Hydraulic 250 kN machine within laboratories) are detailed in this paper. Full article

Postpartum haemorrhage is one of the leading causes of maternal morbidity and mortality worldwide. The early identification of bleeding risk in individual women is crucial for enabling timely interventions and improving patient outcomes.This study aims to evaluate various exploratory and classification methodologies, alongside optimisation strategies, for identifying predictors of postpartum haemorrhage. K-means clustering was employed on a retrospective cohort of patients, incorporating demographic, obstetric, and laboratory variables, to delineate patient profiles and select pertinent features. Initially, a classical logistic regression model, implemented without cross-validation, facilitated the identification of six significant predictors for postpartum haemorrhage: lactate dehydrogenase, urea, platelet count, non-O blood group, gestational age, and first-degree lacerations, all of which are variables routinely available in clinical practice. Furthermore, machine learning algorithms—including stepwise logistic regression, ridge logistic regression, and random forest—were utilised, applying cross-validation to optimise predictive performance and enhance generalisability. Among these methodologies, ridge logistic regression emerged as the most effective model, achieving the following metrics: sensitivity 0.857, specificity 0.875, accuracy 0.871, F1-score 0.759, and AUC 0.907. While machine learning techniques demonstrated superior performance, the integration of classical statistical methods with machine learning approaches provides a robust framework for generating reliable predictions and fostering significant clinical insights. Full article

Compaction quality control in earthworks and pavements still relies mainly on density-based acceptance referenced to laboratory Proctor tests, which are costly, time-consuming, and spatially sparse. Lightweight dynamic cone penetrometer (LDCP) provides rapid indices, such as $q_{d0}$ and $q_{d1}$, yet acceptance thresholds commonly depend on ad hoc, site-specific calibrations. This study develops and validates a supervised machine learning framework that estimates $q_{d0}$, $q_{d1}$, and $Z_c$ directly from readily available soil descriptors (gradation, plasticity/activity, moisture/state variables, and GTR class) using a multi-campaign dataset of $n=360$ observations. While the framework does not remove the need for the standard soil characterization performed during design (e.g., W, ${\gamma }_{d,\mathrm{field}}$, and $R{C}_{\mathrm{SPC}}$), it reduces reliance on additional LDCP calibration campaigns to obtain device-specific reference curves. Models compared under a unified pipeline include regularized linear baselines, support vector regression, Random Forest, XGBoost, and a compact multilayer perceptron (MLP). The evaluation used a fixed 80/20 train–test split with 5-fold cross-validation on the training set and multiple error metrics ($R^2$, RMSE, MAE, and MAPE). Interpretability combined SHAP with permutation importance, 1D partial dependence (PDP), and accumulated local effects (ALE); calibration diagnostics and split-conformal prediction intervals connected the predictions to QA/QC decisions. A naïve GTR-average baseline was added for reference. Computation was lightweight. On the test set, the MLP attained the best accuracy for $q_{d1}$ ($R^2=0.794$, RMSE $=5.866$), with XGBoost close behind ($R^2=0.773$, RMSE $=6.155$). Paired bootstrap contrasts with Holm correction indicated that the MLP–XGBoost difference was not statistically significant. Explanations consistently highlighted density- and moisture-related variables (${\gamma }_{d,\mathrm{field}}$, $R{C}_{\mathrm{SPC}}$, and W) as dominant, with gradation/plasticity contributing second-order adjustments; these attributions are model-based and associational rather than causal. The results support interpretable, computationally efficient surrogates of LDCP indices that can complement density-based acceptance and enable risk-aware QA/QC via conformal prediction intervals. Full article

Designing UHPC beams for shear is challenging because many factors—geometry, concrete strength, fibers, and stirrups—act together. In this study, we compile a large, curated database of laboratory tests and develop machine learning models to predict shear capacity. The best models provide accurate point predictions and, importantly, a 95% prediction band that tells how much uncertainty to expect; in tests, about 95% of results fall inside this band. For day-to-day design, we also offer a short, design-oriented formula with explicit coefficients and variables that can be used in a spreadsheet. Together, these tools let engineers screen options quickly, check designs with an uncertainty margin, and choose a conservative value when needed. The approach is transparent, easy to implement, and aligned with common code variables, so it can support preliminary sizing, verification, and assessment of UHPC members. Full article

Background/Objectives: Non-variceal upper gastrointestinal bleeding (NVUGIB) is associated with considerable morbidity and mortality, particularly in emergency department (ED) settings. While traditional clinical scores such as the Glasgow-Blatchford Score (GBS), AIMS65, and Pre-Endoscopic Rockall are widely used for risk stratification, their accuracy in mortality prediction is limited. This study aimed to evaluate the performance of multiple supervised machine learning (ML) models in predicting 30-day all-cause mortality in NVUGIB and to compare these models with established risk scores. Methods: A retrospective cohort study was conducted on 1233 adult patients with NVUGIB who presented to the ED of a tertiary center between January 2022 and January 2025. Clinical and laboratory data were extracted from electronic records. Seven supervised ML algorithms—logistic regression, ridge regression, support vector machine, random forest, extreme gradient boosting (XGBoost), naïve Bayes, and artificial neural networks—were trained using six feature selection techniques generating 42 distinct models. Performance was assessed using AUROC, F1-score, sensitivity, specificity, and calibration metrics. Traditional scores (GBS, AIMS65, Rockall) were evaluated in parallel. Results: Among the cohort, 96 patients (7.8%) died within 30 days. The best-performing ML model (XGBoost with univariate feature selection) achieved an AUROC > 0.80 and F1-score of 0.909, significantly outperforming all traditional scores (highest AUROC: Rockall, 0.743; p < 0.001). ML models demonstrated higher sensitivity and specificity, with improved calibration. Key predictors consistently included age, comorbidities, hemodynamic parameters, and laboratory markers. The best-performing ML models demonstrated very high apparent AUROC values (up to 0.999 in internal analysis), substantially exceeding conventional scores. These results should be interpreted as apparent performance estimates, likely optimistic in the absence of external validation. Conclusions: While machine-learning models showed markedly higher apparent discrimination than conventional scores, these findings are based on a single-center retrospective dataset and require external multicenter validation before clinical implementation. Full article

Advances in brain–computer interface (BCI) research have explored various strategies for acquiring and processing electroencephalographic (EEG) signals to detect motor imagery (MI) activities. However, the complexity of multichannel clinical systems and processing techniques can limit their accessibility outside specialized centers, where complex setups are not feasible. This paper presents a proof-of-concept prototype of a single-channel EEG acquisition and processing system designed to identify lower-limb motor imagery. The proposed proof-of-concept prototype enables the wireless acquisition of raw EEG values, signal processing using digital filters, and the detection of MI patterns using machine learning algorithms. Experimental validation in a controlled laboratory with participants performing resting, MI, and movement tasks showed that the best performance was obtained by combining Savitzky–Golay filtering with a Random Forest classifier, reaching 87.36% ± 4% accuracy and an F1-score of 87.18% ± 3.8% under five-fold cross-validation. These findings confirm that, despite limited spatial resolution, MI patterns can be detected using appropriate AI-based filtering and classification. The novelty of this work lies in demonstrating that a single-channel, portable EEG prototype can be effectively used for lower-limb MI recognition. The portability and noise resilience achieved with the prototype highlight its potential for research, clinical rehabilitation, and assistive device control in non-specialized environments. Full article

Although current 3D semantic segmentation methods have achieved significant success, they suffer from catastrophic forgetting when confronted with dynamic, open environments. To address this issue, class incremental learning is introduced to update models while maintaining a balance between plasticity and stability. In this work, we propose CosPrompt, a rehearsal-free approach for class incremental semantic segmentation. Specifically, we freeze the prompts for existing classes and incrementally expand and fine-tune the prompts for new classes, thereby generating discriminative and customized features. We employ clamping operations to regulate backward propagation, ensuring smooth training. Furthermore, we utilize the learning without forgetting loss and pseudo-label generation to further mitigate catastrophic forgetting. We conduct comparative and ablation experiments on the S3DIS dataset and ScanNet v2 dataset, demonstrating the effectiveness and feasibility of our method. Full article

Developing climate-resilient and high-quality cotton cultivars remains an urgent challenge, as the key target traits yield, fibre properties, and stress tolerance are highly polygenic and strongly influenced by genotype–environment interactions. Recent advances in chromosome-scale genome assemblies, pan-genomics, and haplotype-resolved resequencing have greatly enhanced the capacity to identify causal variants and recover non-reference alleles linked to fibre development and environmental adaptation. Parallel progress in functional genomics and precision genome editing, particularly CRISPR/Cas, base editing, and prime editing, now enables rapid, heritable modification of candidate loci across the complex tetraploid cotton genome. When integrated with high-throughput phenotyping, genomic selection, and machine learning, these approaches support predictive ideotype design rather than empirical, trial-and-error breeding. Emerging digital agriculture tools, such as digital twins that combine genomic, phenomic, and environmental data layers, allow simulation of ideotype performance and optimisation of trait combinations in silico before field validation. Speed breeding and phenomic selection further shorten generation time and increase selection intensity, bridging the gap between laboratory discovery and field deployment. However, the large-scale implementation of these technologies faces several practical constraints, including high infrastructural costs, limited accessibility for resource-constrained breeding programmes in developing regions, and uneven regulatory acceptance of genome-edited crops. However, reliance on highly targeted genome editing may inadvertently narrow allelic diversity, underscoring the need to integrate these tools with broad germplasm resources and pangenomic insights to sustain long-term adaptability. To realise these opportunities at scale, standardised data frameworks, interoperable phenotyping systems, robust multi-omic integration, and globally harmonised, science-based regulatory pathways are essential. This review synthesises recent progress, highlights case studies in fibre, oil, and stress-resilience engineering, and outlines a roadmap for translating integrative genomics into climate-smart, high-yield cotton breeding programmes. Full article

The growing demand for automated production systems is driving continuous innovation in smart and data-driven manufacturing technologies. In the field of production metrology, the trend is shifting from using measurement laboratories to integrating measurement systems directly into production processes. This has led the Institute of Manufacturing Technology at TU Vienna together with its partners to develop a roughness measurement device that can be directly integrated into machine tools. Building on this foundation, this study tries to find applications beyond mere surface roughness assessment and demonstrates how the device could be applied in broader contexts of manufacturing process monitoring. By linking surface measurements with tool wear monitoring, the study establishes a correlation between surface roughness and wear progression of indexable inserts in milling. It demonstrates how in situ data can support predictive maintenance and the real-time adjustment of cutting parameters. This represents a first step toward integrating in situ metrology into closed-loop control in machining. The experimental setup followed ISO 8688-1 guidelines for tool life testing. Indexable inserts were operated throughout their entire service life while surface roughness was continuously recorded. In parallel, cutting edge conditions were documented at defined intervals using focus variation microscopy. The results show a consistent three-phase pattern: initially stable roughness, followed by a steady increase due to flank wear, and an abrupt decrease in roughness linked to edge chipping. These findings confirm the potential of integrated roughness measurement for condition-based monitoring and the development of adaptive machining strategies. Full article

Background: Hepatocellular carcinoma (HCC) patients without microvascular invasion (MVI) face significant postoperative early recurrence (ER) risks, yet prognostic determinants remain understudied. Existing models often rely on linear assumptions. This study aimed to develop and validate an interpretable machine learning model using routine clinical parameters to predict early recurrence (ER) in MVI-negative HCC patients. Methods: We retrospectively analyzed 578 MVI-negative HCC patients undergoing radical resection. Seven machine learning (ML) algorithms were systematically benchmarked using clinical/laboratory/imaging features optimized via recursive feature elimination (RFE) and hyperparameter tuning. Model interpretability was achieved via SHapley Additive exPlanations (SHAP). Results: The CatBoost model demonstrated superior performance (AUC: 0.7957, Accuracy: 0.7290). SHAP analysis identified key predictors: tumor capsule absence, elevated HBV-DNA and CA125 levels, larger tumor diameter, and lower body weight significantly increased ER risk. Individualized SHAP force plots enhanced clinical interpretability. Conclusions: The CatBoost model exhibits robust predictive performance for ER in MVI-negative HCC, offering a clinically interpretable tool for personalized risk stratification and optimization of postoperative management strategies. Full article

Background: Microvascular invasion (MVI) is a critical prognostic factor in hepatocellular carcinoma (HCC), but preoperative three-class prediction remains challenging. Radiomics and clinical biomarkers may enable more accurate and individualized assessment. Aim: The aim of this study was to develop and validate a Transformer-based deep learning framework that integrates radiomic and clinical features for direct three-class MVI classification in HCC patients. Methods: This retrospective study included 437 patients with pathologically confirmed hepatocellular carcinoma (HCC) and microvascular invasion (MVI) status from two campuses of a single institution. Patients from Hospital A (n = 305) were randomly divided into training and internal test cohorts, while patients from Hospital B (n = 132) were used as an independent external validation cohort. Radiomic features were extracted from preoperative Gd-BOPTA-enhanced MRI, and clinical laboratory data were collected. A two-stage feature selection strategy, combining univariate statistical testing and recursive feature elimination, was applied. A Transformer-based model was built to classify three MVI categories (M0, M1, M2), and its performance was evaluated in both the internal test cohort and the external validation cohort. Results were compared with those from traditional machine learning models, including Random Forest, Logistic Regression, XGBoost, and LightGBM. Results: On the internal test set (n = 76, Hospital A), the model achieved an accuracy of 0.733 (95% CI: 0.64–0.83), a weighted F1-score of 0.733, and a macro-average AUC of 0.880 (95% CI: 0.807–0.953). The sensitivity and specificity for M1 were 0.56 (95% CI: 0.31–0.78) and 0.86 (95% CI: 0.74–0.94), respectively; for high-risk M2 cases, the sensitivity was 0.73 (95% CI: 0.64–0.81) and the specificity was 0.91 (95% CI: 0.85–0.96). On the external validation set (n = 132, Hospital B), performance remained stable with an accuracy of 0.758, a weighted F1-score of 0.768, and a macro-average AUC of 0.886 (95% CI: 0.833–0.940). Conclusions: This Transformer-based model enables accurate and objective three-class MVI prediction using multi-modal features, supporting individualized surgical planning and improved clinical outcomes. In particular, the ability to preoperatively identify high-risk M2 patients may inform surgical margin design, guide adjuvant therapy strategies, and influence liver transplantation eligibility. Full article

Wearable sensors have become valuable tools for assessing gait in both laboratory and free-living environments. However, detection of walking in free-living environments remains challenging, especially in clinical populations. Machine learning models may offer more robust gait identification, but most are trained on healthy participants, limiting their generalizability to other populations. To extend a previously validated machine learning model, an updated model was trained using an open dataset (PAMAP2), before progressively including training datasets with additional healthy participants and a clinical osteoarthritis population. The performance of the model in identifying walking was also evaluated using a frequency-based gait detection algorithm. The results showed that the model trained with all three datasets performed best in terms of activity classification, ultimately achieving a high accuracy of 96% on held-out test data. The model generally performed on par with the heuristic, frequency-based method for walking bout identification. However, for patients with slower gait speeds (<0.8 m/s), the machine learning model maintained high recall (>0.89), while the heuristic method performed poorly, with recall as low as 0.38. This study demonstrates the enhancement of existing model architectures by training with diverse datasets, highlighting the importance of dataset diversity when developing more robust models for clinical applications. Full article