Search Results (450)

Type 2 diabetes mellitus (T2DM) is a growing global health challenge, expected to affect over 600 million people by 2045. The discovery of new antidiabetic agents remains resource-intensive, motivating the use of machine learning (ML) for virtual screening based on molecular structure. In this study, we developed a predictive pipeline integrating two distinct descriptor types: high-dimensional numerical features from the Mordred library (>1800 2D/3D descriptors) and categorical ontological annotations from the ClassyFire and ChEBI systems. These encode hierarchical chemical classifications and functional group labels. The dataset included 45 active compounds and thousands of inactive molecules, depending on the descriptor system. To address class imbalance, we applied SMOTE and created balanced training and test sets while preserving independent validation sets. Thirteen ML models—including regression, SVM, naive Bayes, decision trees, ensemble methods, and others—were trained using stratified 12-fold cross-validation and evaluated across training, test, and validation. Ridge Regression showed the best generalization (MCC = 0.814), with Gradient Boosting following (MCC = 0.570). Feature importance analysis highlighted the complementary nature of the descriptors: Ridge Regression emphasized ClassyFire taxonomies such as CHEMONTID:0000229 and CHEBI:35622, while Mordred-based models (e.g., Random Forest) prioritized structural and electronic features like MAXsssCH and ETA_dEpsilon_D. This study is the first to systematically integrate and compare structural and ontological descriptors for antidiabetic compound prediction. The framework offers a scalable and interpretable approach to virtual screening and can be extended to other therapeutic domains to accelerate early-stage drug discovery. Full article

Early plant disease detection is crucial for sustainable crop production and food security. Stem rust, caused by Puccinia graminis f. sp. tritici, poses a major threat to wheat and barley. This study evaluates the feasibility of using hyperspectral imaging and machine learning for early detection of stem rust and examines the cross-crop transferability of diagnostic models. Hyperspectral datasets of wheat (Triticum aestivum L.) and barley (Hordeum vulgare L.) were collected under controlled conditions, before visible symptoms appeared. Multi-stage preprocessing, including spectral normalization and standardization, was applied to enhance data quality. Feature engineering focused on spectral curve morphology using first-order derivatives, categorical transformations, and extrema-based descriptors. Models based on Support Vector Machines, Logistic Regression, and Light Gradient Boosting Machine were optimized through Bayesian search. The best-performing feature set achieved F1-scores up to 0.962 on wheat and 0.94 on barley. Cross-crop transferability was evaluated using zero-shot cross-domain validation. High model transferability was confirmed, with F1 > 0.94 and minimal false negatives (<2%), indicating the universality of spectral patterns of stem rust. Experiments were conducted under controlled laboratory conditions; therefore, direct field transferability may be limited. These findings demonstrate that hyperspectral imaging with robust preprocessing and feature engineering enables early diagnostics of rust diseases in cereal crops. Full article

PostCOVID-19 is a condition affecting approximately 10% of individuals infected with SARS-CoV-2, presenting significant challenges in diagnosis and clinical management. Portable neuroimaging techniques, such as functional near-infrared spectroscopy (fNIRS), offer real-time insights into cerebral hemodynamics and represent a promising tool for studying postCOVID-19 in naturalistic settings. This study investigates the integration of fNIRS with machine learning to identify neural correlates of postCOVID-19. A total of six machine learning classifiers—Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNNs), XGBoost, Logistic Regression, and Multi-Layer Perceptron (MLP)—were evaluated using a stratified subject-aware cross-validation scheme on a dataset comprising 29,737 time-series samples from 37 participants (9 postCOVID-19, 28 controls). Four different feature representation strategies were compared: raw time-series, PCA-based dimensionality reduction, statistical feature extraction, and a hybrid approach that combines time-series and statistical descriptors. Among these, the hybrid representation demonstrated the highest discriminative performance. The SVM classifier trained on hybrid features achieved strong discrimination ($\mathit{ROC}\text{-}\mathit{AUC}$ = 0.909) under subject-aware CV5; at the default threshold, $\mathit{Sensitivity}$ was moderate and $\mathit{Specificity}$ was high, outperforming all other methods. In contrast, models trained on statistical features alone exhibited limited $\mathit{Sensitivity}$ despite high $\mathit{Specificity}$. These findings highlight the importance of temporal information in the fNIRS signal and support the potential of machine learning combined with portable neuroimaging for postCOVID-19 identification. This approach may contribute to the development of non-invasive diagnostic tools to support individualized treatment and longitudinal monitoring of patients with persistent neurological symptoms. Full article

Previous researches have proved that the synchronized use of inexpensive RGB images, image processing, and machine learning (ML) can accurately identify crop stress. Four Machine Learning Image Modules (MLIMs) were developed to enable the rapid and cost-effective identification of sugar beet stresses caused by water and/or nitrogen deficiencies. RGB images representing stressed and non-stressed crops were used in the analysis. To improve robustness, data augmentation was applied, generating six variations on each image and expanding the dataset from 150 to 900 images for training and testing. Each MLIM was trained and tested using 54 combinations derived from nine canopy and RGB-based input features and six ML algorithms. The most accurate MLIM used RGB bands as inputs to a Multilayer Perceptron, achieving 96.67% accuracy for overall stress detection, and 95.93% and 94.44% for water and nitrogen stress identification, respectively. A Random Forest model, using only the green band, achieved 92.22% accuracy for stress detection while requiring only one-fourth the computation time. For specific stresses, a Random Forest (RF) model using a Scale-Invariant Feature Transform descriptor (SIFT) achieved 93.33% for water stress, while RF with RGB bands and canopy cover reached 85.56% for nitrogen stress. To address the trade-off between accuracy and computational cost, a bargaining theory-based framework was applied. This approach identified optimal MLIMs that balance performance and execution efficiency. Full article

Data augmentation is crucial for electric motor fault diagnosis and lifetime prediction. However, the diversity of operating conditions and the challenge of augmenting small datasets often limit existing models. To address this, we propose an enhanced TimeGAN framework that couples the original architecture with transformer modules to jointly exploit time- and frequency-domain information to improve the fidelity of synthetic motor signals. The method fuses raw waveforms, envelope features, and instantaneous phase-change cues to strengthen temporal representation learning. The generator further incorporates frequency-domain descriptors and adaptively balances time–frequency contributions through learnable weighting, thereby improving generative performance. In addition, a state-conditioning mechanism (via explicit state annotations) enables controlled synthesis across distinct operating states. Comprehensive evaluations—including PCA and t-SNE visualizations, distance metrics such as DTW and FID, and downstream classifier tests—demonstrate strong adaptability and robustness on both public and in-house datasets, substantially enhancing the quality of generated time series. Full article

Parkinson’s disease (PD) is one of the fastest-growing neurodegenerative disorders; its cardinal motor signs—tremor, bradykinesia, and rigidity—substantially impair quality of life. Conventional clinician-rated scales can be subjective and exhibit limited interrater reliability, underscoring the need for objective and reliable quantification. We present an integrated evaluation framework that leverages surface electromyography (sEMG) with multimodal sensing. For representation learning, we combine time–frequency descriptors with Mini-ROCKET features. Grading is performed by an sEMG-based Unified Parkinson’s Disease Rating Scale (UPDRS) model (LDA-SV) that produces per-segment probabilities for ordinal scores (0–3) and aggregates them via soft voting to assign item-level ratings. Participants completed a standardized protocol spanning gait, seated rest, and upper-limb tasks (forearm pronation–supination, finger-to-nose, fist clench, and thumb–index pinch). Using the aforementioned dataset, we report task-wise performance with 95% confidence intervals and compare the proposed model against CNN, LSTM, and InceptionTime using McNemar tests and log-odds ratios. The results indicate that the proposed model outperforms the three baseline models overall. These findings demonstrate the effectiveness and feasibility of the proposed approach, suggesting a viable pathway for the objective quantification of PD motor symptoms and facilitating broader clinical adoption of sEMG in diagnosis and treatment. Full article

The field of autonomous driving has seen continuous advances, yet achieving higher levels of automation in real-world applications remains challenging. A critical requirement for autonomous navigation is accurate map construction, particularly in novel and unstructured environments. In recent years, Simultaneous Localization and Mapping (SLAM) has evolved to support diverse sensor modalities, with some implementations incorporating machine learning to improve performance. However, these approaches often demand substantial computational resources. The key challenge lies in achieving efficiency within resource-constrained environments while minimizing errors that could degrade downstream tasks. This paper presents an enhanced ORB-SLAM2 (Oriented FAST and Rotated BRIEF Simultaneous Localization and Mapping, version 2) algorithm implemented on a Raspberry Pi 3 (ARM A53 CPU) to improve mapping performance under limited computational resources. ORB-SLAM2 comprises four main stages: Tracking, Local Mapping, Loop Closing, and Full Bundle Adjustment (BA). The proposed improvements include employing a more efficient feature descriptor to increase stereo feature-matching rates and optimizing loop-closing parameters to reduce accumulated errors. Experimental results demonstrate that the proposed system achieves notable improvements on the Raspberry Pi 3 platform. For monocular SLAM, RMSE is reduced by 18.11%, mean error by 22.97%, median error by 29.41%, and maximum error by 17.18%. For stereo SLAM, RMSE decreases by 0.30% and mean error by 0.38%. Furthermore, the ROS topic frequency stabilizes at 10 Hz, with quad-core CPU utilization averaging approximately 90%. These results indicate that the system satisfies real-time requirements while maintaining a balanced trade-off between accuracy and computational efficiency under resource constraints. Full article

Efficiently identifying suitable solvents for active pharmaceutical ingredients (APIs) is critical in drug formulation, yet the vast number of possible solvent-solute combinations presents a significant experimental challenge. This study addresses this by developing a robust machine learning (ML) model for accurately predicting the solubility of three phenolic acids (syringic, p-coumaric, and caffeic) in various deep eutectic solvents (DESs), integrating both experimental and computational investigations. Measured solubility data showed that the choline chloride combined with triethylene glycol in a 1:2 molar ratio was the most efficient system for the dissolution of the studied APIs. Different ML models, utilizing nu-Support Vector Regression (nuSVR) as the core regressor and based on descriptor sets derived from COSMO-RS (Conductor-like Screening Model for Real Solvents) computations, were systematically evaluated. A novel methodology termed DOO-IT (Dual-Objective Optimization with ITerative feature pruning) was employed to address the common challenges of model development with limited, high-value datasets. The final optimal 10-descriptor nuSVR model, selected from an exhaustive, multi-run search, demonstrated outstanding predictive power, offering a highly reliable computational tool for guiding experimental screening, significantly accelerating the exploration of DES-based formulations. This research also provides a strong foundation for future machine learning-guided discovery of chemicals, offering an effective and transferable framework for developing QSPR models for various chemical systems. Full article

In surgical navigation-assisted pedicle screw fixation, cross-source pre- and intra-operative point clouds registration faces challenges like significant initial pose differences and low overlapping ratio. Classical algorithms based on feature descriptor have high computational complexity and are less robust to noise, leading to a decrease in accuracy and navigation performance. To address these problems, this paper proposes a coarse-to-fine registration framework. In the coarse registration stage, a Point Matching algorithm based on Curvature Feature Learning (CFL-PM) is proposed. Through CFL-PM and Farthest Point Sampling (FPS), the coarse registration of overlapping regions between the two point clouds is achieved. In the fine registration stage, the Iterative Closest Point (ICP) is used for further optimization. The proposed method effectively addresses the challenges of noise, initial pose and low overlapping ratio. In noise-free point cloud registration experiments, the average rotation and translation errors reached 0.34° and 0.27 mm. Under noisy conditions, the average rotation error of the coarse registration is 7.28°, and the average translation error is 9.08 mm. Experiments on pre- and intra-operative point cloud datasets demonstrate the proposed algorithm outperforms the compared algorithms in registration accuracy, speed, and robustness. Therefore, the proposed method can achieve the precise alignment of the surgical navigation-assisted pedicle screw fixation. Full article

A crucial task in industrial quality control, especially in the food and agriculture sectors, is the quick and precise estimation of an object’s volume. This study combines cutting-edge machine learning and computer vision techniques to provide a comprehensive, non-destructive method for predicting orange volume. We created a reliable pipeline that employs top and side views of every orange to estimate four important dimensions using a calibrated marker. These dimensions are then fed into a machine learning model that has been fine-tuned. Our method uses a range of engineered features, such as complex surface-area-to-volume ratios and new shape-based descriptors, to go beyond basic geometric formulas. Based on a dataset of 150 unique oranges, we show that the Stacking Regressor performs significantly better than other single-model benchmarks, including the highly tuned LightGBM model, achieving an $R^2$ score of 0.971. Because of its reliance on basic physical characteristics, the method is extremely resilient to the inherent variability in fruit and may be used with a variety of produce types. Because it allows for the real-time calculation of density (mass over volume) for automated defect detection and quality grading, this solution is directly applicable to a factory sorting environment. Full article

Rare-earth transition metal-based double perovskite ceramics E₂TMO₆ (where E = rare-earth metals, T = transition metals, and M = metal) have received impressive attention lately for cryogenic applications as a result of their intrinsic physical features such as multiferroicity, dielectric features, and adjustable magnetic transition temperature. However, determination and enhancement of magnetic transition temperature of E₂TMO₆ ceramic are subject to experimental procedures and processes with a significant degree of difficulties and cumbersomeness. This work proposes an extreme learning machine (ELM)-based intelligent method of determining magnetic transition temperature of E₂TMO₆ ceramics with activation function sigmoid (SM) and sine (SE) at varying magnetic field. The outcomes of the SE-ELM and SM-ELM models were compared with genetically optimized support vector regression (GEN-SVR) predictive models using RMSE, CC, and MAE metrics. Using the testing samples of E₂TMO₆ ceramics, SE-ELM predictive model outperforms GEN-SVR with a superiority of 6.3% (using RMSE metric) and 15.7% (using MAE metric). The SE-ELM predictive model further outperforms the SM-ELM model, with an improvement of 5.3%, using CC computed with training ceramic samples. The simplicity of the employed descriptors, coupled with the outstanding performance of the developed predictive models, would potentially strengthen E₂TMO₆ ceramics exploration for low-temperature cryogenic applications and circumvent energy challenges in different sectors. Full article

Automated quality control is critical in modern manufacturing, especially for metallic cast components, where fast and accurate surface defect detection is required. This study evaluates classical Machine Learning (ML) algorithms using extracted statistical parameters and deep learning (DL) architectures including ResNet50, Capsule Networks, and a 3D Convolutional Neural Network (CNN3D) using 3D image inputs. Using the Dataset Original, ML models with the selected parameters achieved high performance: RF reached 99.4 ± 0.2% precision and 99.4 ± 0.2% sensitivity, GB 96.0 ± 0.2% precision and 96.0 ± 0.2% sensitivity. ResNet50 trained with extracted parameters reached 98.0 ± 1.5% accuracy and 98.2 ± 1.7% F1-score. Capsule-based architectures achieved the best results, with ConvCapsuleLayer reaching 98.7 ± 0.2% accuracy and 100.0 ± 0.0% precision for the normal class, and 98.9 ± 0.2% F1-score for the affected class. CNN3D applied on 3D image inputs reached 88.61 ± 1.01% accuracy and 90.14 ± 0.95% F1-score. Using the Dataset Expanded with ML and PCA-selected features, Random Forest achieved 99.4 ± 0.2% precision and 99.4 ± 0.2% sensitivity, K-Nearest Neighbors 99.2 ± 0.0% precision and 99.2 ± 0.0% sensitivity, and SVM 99.2 ± 0.0% precision and 99.2 ± 0.0% sensitivity, demonstrating consistent high performance. All models were evaluated using repeated train-test splits to calculate averages of standard metrics (accuracy, precision, recall, F1-score), and processing times were measured, showing very low per-image execution times (as low as $3.69\times {10}^{-4}$ s/image), supporting potential real-time industrial application. These results indicate that combining statistical descriptors with ML and DL architectures provides a robust and scalable solution for automated, non-destructive surface defect detection, with high accuracy and reliability across both the original and expanded datasets. Full article

Non-small cell lung cancer (NSCLC) is the leading cause of cancer-related mortality worldwide. Accurate determination of epidermal growth factor receptor (EGFR) mutation status is essential for selecting patients eligible for tyrosine kinase inhibitors (TKIs). However, invasive genotyping is often limited by tissue accessibility and sample quality. This study presents a non-invasive machine learning model combining clinical data, CT morphological features, and radiomic descriptors to predict EGFR mutation status. A retrospective cohort of 138 patients with confirmed EGFR status and pre-treatment CT scans was analyzed. Radiomic features were extracted with PyRadiomics, and feature selection applied mutual information, Spearman correlation, and wrapper-based methods. Five Random Forest models were trained with different feature sets. The best-performing model, based on 11 selected variables, achieved an AUC of 0.91 (95% CI: 0.81–1.00) under stratified five-fold cross-validation, with an accuracy of 0.88 ± 0.03. Subgroup analysis showed that EGFR-WT had a performance of precision 0.93 ± 0.04, recall 0.92 ± 0.03, F1-score 0.91 ± 0.02, and EGFR-Mutant had a performance of precision 0.76 ± 0.05, recall 0.71 ± 0.05, F1-score 0.68 ± 0.04. SHapley Additive exPlanations (SHAP) analysis identified tobacco use, enhancement pattern, and gray-level-zone entropy as key predictors. Decision curve analysis confirmed clinical utility, supporting its role as a non-invasive tool for EGFR-screening. Full article

Line feature matching is a fundamental and extensively studied subject in the fields of photogrammetry and computer vision. Traditional methods, which rely on handcrafted descriptors and distance-based filtering outliers, frequently encounter challenges related to robustness and a high incidence of outliers. While some approaches leverage point features to assist line feature matching by establishing the invariant geometric constraints between points and lines, this typically results in a considerable computational load. In order to overcome these limitations, we introduce a novel Bayesian posterior probability framework for line matching that incorporates three geometric constraints: the distance between line feature endpoints, midpoint distance, and angular consistency. Our approach initially characterizes inter-image geometric relationships using Fourier representation. Subsequently, we formulate the posterior probability distributions for the distance constraint and the uniform distribution based on the constraint of angular consistency. By calculating the joint probability distribution under three geometric constraints, robust line feature matches are iteratively optimized through the Expectation–Maximization (EM) algorithm. Comprehensive experiments confirm the effectiveness of our approach: (i) it outperforms state-of-the-art (including deep learning-based) algorithms in match count and accuracy across common scenarios; (ii) it exhibits superior robustness to rotation, illumination variation, and motion blur compared to descriptor-based methods; and (iii) it notably reduces computational overhead in comparison to algorithms that involve point-assisted line matching. Full article

Background/Objectives: The YSK olfactory function (YOF) test is a culturally adapted psychophysical tool that assesses threshold, discrimination, and identification. This study evaluated whether functional near-infrared spectroscopy (fNIRS) synchronized with routine YOF testing, combined with machine learning, can predict YOF subdomain performance in healthy adults, providing an objective neural correlate to complement behavioral testing. Methods: In this prospective diagnostic-accuracy (feasibility/validation) study in healthy adults with algorithm development, 100 healthy adults completed the YOF test while undergoing prefrontal/orbitofrontal fNIRS during odor blocks. Feature sets from ΔHbO/ΔHbR included time-domain descriptors, complexity (Lempel–Ziv), and information-theoretic measures (mutual information); the identification task used a hybrid attention–CNN. Separate models were developed for threshold (binary classification), discrimination (binary classification), and identification (binary classification). Performance was summarized with accuracy, area under the curve (AUC), F1-score, and (where applicable) sensitivity/specificity, using participant-level cross-validation. Results: The threshold classifier achieved accuracy 0.86, AUC 0.86, and F1 0.86, indicating strong discrimination of correct vs. incorrect threshold responses. The discrimination model yielded accuracy 0.75, AUC 0.76, and F1 0.75. The identification model (attention–convolutional neural network [CNN]) achieved accuracy 0.88, sensitivity 0.86, specificity 0.91, and F1 0.88. Feature-attribution (e.g., SHapley Additive exPlanations [SHAP]) provided interpretable links between fNIRS features and task performance for threshold and discrimination. Conclusions: Olfactory-evoked fNIRS signals can accurately predict YOF subdomain performance in healthy adults, supporting the feasibility of non-invasive, portable, near–real-time olfactory monitoring. These findings are preliminary and not generalizable to clinical populations; external validation in diverse cohorts is warranted. The approach clarifies the scientific essence of the method by (i) aligning psychophysical outcomes with objective hemodynamic signatures and (ii) introducing a feature-rich modeling pipeline (ΔHbO/ΔHbR + Lempel–Ziv complexity/mutual information; attention–CNN) that advances prior work. Full article