Early-warning systems (EWS) in learning analytics are increasingly used to identify learners at risk of course withdrawal, but their deployment-critical properties are often under-reported once predicted scores are converted into intervention policies. This study develops a deployment-oriented evaluation protocol for course-withdrawal risk using the Open University Learning Analytics Dataset (OULAD). An early-window feature set was constructed from the first four weeks of learner activity and evaluated under a group-wise train-test split by course presentation. Multiple classifiers were benchmarked, including logistic regression, histogram-based gradient boosting (HGB), random forest, support vector machine, AdaBoost, K-nearest neighbors, XGBoost, LightGBM, and CatBoost. A calibrated HGB model was then retained as the main probabilistic model for downstream analyses of probability reliability, threshold sensitivity, subgroup fairness with bootstrap uncertainty, selective prediction, and capacity-based Top-x% alerting. Several tree-based and boosting models achieved comparable held-out discrimination, while calibrated HGB remained competitive across classification, ranking, and probability-reliability metrics. Threshold choices substantially changed the precision-recall balance, indicating that operating points should be treated as policy choices rather than universal defaults. Fairness audits showed policy-dependent observed group-level differences, especially in alert rates for disability status, although several subgroup error-rate and positive predictive value (PPV) differences remained uncertain. Selective prediction reduced risk on accepted cases as coverage decreased, whereas Top-x% alerting fixed outreach volume and made workload-effectiveness trade-offs explicit. Robustness analyses supported the 28-day window as a practical early-warning compromise and showed that absolute PPV values varied across held-out course-presentation splits. The findings suggest that EWS should be evaluated as policy-linked decision systems, integrating model benchmarking, calibration, fairness uncertainty, and capacity-aware decision rules before deployment.

Incentive-based demand response is an important flexibility resource for power systems with high-renewable energy penetration. However, practical incentive allocation depends not only on flexible capacity and user response uncertainty, but also on the credibility of customer baseline load (CBL), which directly affects response measurement, verification, and incentive settlement. To address this issue, this paper proposes a constrained reinforcement learning method with customer baseline credibility for dynamic resource allocation in incentive-based demand response. Based on user-side load measurements and demand response event records, the proposed framework evaluates user resources using flexible capacity, response reliability, response cost, and CBL credibility. The CBL credibility score reflects the measurement quality of the delivered response and is used as a pre-event allocation factor. Users are then grouped into different resource levels, and a group-level reinforcement learning agent dynamically determines incentive multipliers and response task allocation ratios. To improve feasibility, an action correction module revises raw policy outputs under budget, price, response capacity, and CBL risk constraints before implementation. Case studies are conducted using public industrial demand response measurements and open electricity-system time-series data. The results show that the proposed CBL-CRL method reduces the normalized total operating cost to 0.897, reduces the response tracking error to 0.108, and lowers CBL risk exposure to 0.087 under the normal scenario. Relative to the No-DR reference, CBL-CRL reduces the normalized total operating cost by 10.3 percent. Compared with MAPPO, the strongest learning-based baseline, CBL-CRL reduces the response tracking error by 10.7 percent and the CBL risk exposure by 40.8 percent, while maintaining the same renewable accommodation rate of 0.970. Compared with rule-based and learning-based baselines, CBL-CRL achieves a better balance between operational performance, incentive efficiency, action feasibility, and baseline-related settlement reliability. The results demonstrate that CBL credibility should not only be used for post-event settlement, but can also serve as an effective pre-event resource allocation factor for measurement-driven demand response programs.

Background: Large language models (LLMs) are increasingly used by patients for obtaining medical information; however, concerns remain regarding their clinical safety, reliability, and appropriateness of patient guidance. Evidence evaluating LLM performance in hemorrhoid-related patient questions remains limited. Objective: To compare the clinical accuracy, safety, and overall clinical adequacy of responses generated by ChatGPT, Gemini, and DeepSeek to hemorrhoid-related patient questions. Methods: In this cross-sectional comparative study, 25 hemorrhoid-related patient questions were developed and categorized into three predefined subgroups: basic informational questions, clinically significant scenarios, and misleading/risky patient statements. Responses generated by ChatGPT (GPT-5.3), Gemini (3.1), and DeepSeek (R1) were evaluated by two experienced surgeons using a consensus-based expert assessment approach and a structured 5-point scoring system assessing clinical accuracy, safety, appropriateness of patient guidance, and overall clinical adequacy. Critical errors and qualitative communication characteristics were also analyzed. Friedman and post hoc Conover tests with Bonferroni correction were used for statistical comparisons. Results: Overall response quality differed significantly among models (χ[2](2) = 29.119, p < 0.001, Kendall's W = 0.582). ChatGPT achieved the highest overall scores (5.00 ± 0.00), followed by Gemini (4.80 ± 0.41) and DeepSeek (4.12 ± 0.67). Significant differences were primarily observed between DeepSeek and the other models, whereas ChatGPT and Gemini showed comparable performance. Model divergence became more pronounced in clinically significant scenarios involving alarm symptoms, rectal bleeding, persistent symptoms, and acute anorectal pain. No model generated directly harmful medical recommendations or explicit guidance likely to result in substantial diagnostic delay. However, qualitative assessment demonstrated differences in communication style and risk communication. ChatGPT generally produced more balanced and context-appropriate responses, Gemini generated more explanatory responses, whereas DeepSeek showed a tendency toward disproportionately urgent or alarmist language in some higher-risk scenarios. Conclusions: Large language models demonstrated generally high clinical accuracy in answering hemorrhoid-related patient questions; however, notable model-specific differences were observed in clinical guidance, communication style, and risk communication, particularly in high-risk clinical scenarios. These findings suggest that LLMs may serve as useful supportive tools for patient education and health information delivery, although they should currently be regarded as systems that support rather than replace human clinical judgment.

Balance remains a critical element in the evaluation and management of sports injury, with recent recommendations to integrate dual-task cognitive exercises into balancing to further challenge the body's vestibular and somatosensory systems. Dual-task balancing has yet to be examined in a sample of para-athletes. The purpose of this study was to examine the effects of a dual-task cognitive exercise on unsupported, seated balance in a sample of healthy, collegiate para-athletes. A secondary objective was to investigate the role of sex, concussion history, and functional classification.

Age-related declines in neuromuscular and sensory systems substantially increase fall risk and impair independence in older adults. Exercise interventions improve balance and gait; however, the maintenance of these training-induced gains over time remains uncertain, as most benefits may diminish within months after training ceases. This study determined whether an 8-week Dynamic Neuromuscular Stabilization (DNS) programme produces post-intervention and short-term follow-up improvements in balance, gait speed, fear of falling, and health-related quality of life in community-dwelling older adults. In this assessor-blinded randomised controlled trial, 44 older adults aged 60-75 years were randomly allocated to either supervised DNS (totaling 24 sessions: three sessions per week for 8 weeks, 40-50 min per session) or a usual-activity control group. Outcomes were measured at baseline, immediately post-intervention, and at 2-month follow-up using validated clinical instruments. Thirty-nine participants completed the trial (DNS group, n = 19; control group, n = 20). Mixed-design ANOVA showed significant group × time interactions for all outcomes (all P < 0.001). At post-test, the DNS group exhibited large effect-size improvements compared with controls: static balance errors decreased by 58-74% (Cohen's d = 2-2.65), dynamic balance scores increased by 36% (d = 2.05), gait speed increased by 28% (d = 1.54), fear of falling decreased by 42% (d = 1.92), and physical and mental health-related quality of life (HRQOL) components improved substantially (d = 2.20-2.28). At 2-month follow-up, these improvements were largely maintained without significant deterioration within the follow-up period (P > 0.05). The control group exhibited no meaningful changes. An 8-week DNS intervention yields robust, clinically meaningful, and sustained short-term improvements across multiple domains related to fall risk in older adults. These short-term retained adaptations position DNS as an effective neurophysiologically based approach for fall prevention within the observed follow-up period.Trial registration: RTC, prospectively registered in the Clinical Trial Registry (UMIN000055127) on 29/12/2024.

Brinjal (eggplant) is a critical crop in South Asia, especially in Bangladesh, but its production is drastically affected by numerous diseases that inhibit yield and quality. Manual diagnosis of disease is time-consuming, subjective, and prone to errors, necessitating automated, scalable technology. To address these issues, this paper proposes PD-ViCo, a lightweight, efficient transformer-based model for brinjal fruit disease classification using Simple Vision Transformer (ViT) with Patch Dropout and Contrastive Captioner (CoCa) methods. One new dataset of 1,823 field-harvested brinjal images encompassing five disease classes including Phomopsis Blight, Fruit and Shoot Borer, Fruit Cracking, Wet Rot, and Healthy samples were prepared through real-world agricultural data collection from Bangladesh. The approach includes extensive preprocessing, class balancing (under-sampling/oversampling), and resilient augmentation methods. The PD-ViCo model significantly improves classification performance under data imbalance with patch dropout regularization and CoCa-style aggregation, resulting in better generalization and robustness. On a range of imbalanced, under-sampled, and oversampled datasets, PD-ViCo achieved a classification accuracy of 99.12% and F1-score of 97.76%, outperforming both ViT and Swin Transformer across all key evaluation metrics. Explainability was also applied using Grad-CAM and Grad-CAM + + , generating visual explanations of model decisions and maintaining conformity to disease-affected regions in the images. These visualizations ensure the credibility of the model and its usability for real agricultural conditions. This study demonstrates that PD-ViCo is a highly accurate, interpretable, and lightweight model for multi-class brinjal disease diagnosis. Not only does it advance state-of-the-art in agricultural AI, but it also provides a valuable dataset and an understandable decision-making protocol that can be applied directly by farmers, agronomists, and agricultural technologists.

Background/Objectives: Brain magnetic resonance imaging (MRI) is an important imaging modality for assessing neurological disorders. However, automatic multi-class MRI classification remains challenging because of visual similarity between disease categories, heterogeneous pathological patterns, class imbalance, and the need for reliable confidence estimation. This study aims to develop a comprehensive and well-calibrated deep learning framework for image-level brain MRI classification across multiple neurological categories. Methods: This paper introduces a new deep learning framework, MCND-ComputeNet++, for brain MRI classification into eight image-level categories using the MCND dataset, which comprises 16,400 two-dimensional brain MRI images belonging to eight diagnostic categories: AD-MildDemented, AD-ModerateDemented, AD-VeryMildDemented, BT-glioma, BT-meningioma, BT-pituitary, MS, and Normal. The proposed model uses a single pretrained EfficientNetV2-S backbone to extract hierarchical feature maps from three intermediate stages. These multi-level features are projected into a common latent space, spatially aligned, adaptively fused through learnable gated multi-scale fusion, further refined using convolutional processing, and aggregated using spatial attention pooling before classification. The training strategy combines class-balanced focal loss with label smoothing, MixUp/CutMix regularization, exponential moving average weight smoothing, warmup cosine learning-rate scheduling, temperature scaling, and test-time augmentation to improve generalization and calibration. The framework was evaluated using accuracy, precision, recall, macro-F1, macro-AUC, macro-average precision, expected calibration error, Brier score, bootstrap confidence intervals, ablation analysis, McNemar testing, and comparisons against standard pretrained baseline models. Results: MCND-ComputeNet++ achieved mean accuracy, macro-F1, macro-AUC, and macro-average precision values of 0.9738, 0.9771, 0.9993, and 0.9971, respectively, with narrow bootstrap confidence intervals indicating stable image-level performance. These findings should be interpreted as image-level/slice-level performance on MCND, because patient-level identifiers and subject-wise splitting were not available. These results outperformed most evaluated baselines, including ResNet50, DenseNet121, EfficientNetB0, EfficientNetV2-S with a standard classifier, Swin-Tiny, and ConvNeXt-Tiny, across several discrimination and calibration metrics. Compared with ConvNeXt-Tiny, the proposed model achieved higher macro-AUC and macro-average precision, together with a lower ECE and Brier score, suggesting improved image-level discrimination and confidence reliability. Compared with the EfficientNetV2-S standard classifier, accuracy increased from 0.9308 to 0.9738, while the Brier score decreased from 0.1045 to 0.0400. Conclusions: The results suggest that MCND-ComputeNet++ is a promising image-level brain MRI classification framework for the eight MCND categories. The proposed model integrates hierarchical feature extraction, shared latent projection, gated multi-scale fusion, convolutional refinement, spatial attention pooling, and calibrated inference within a unified architecture. However, because the current evaluation was conducted at the image/slice level without available patient-level identifiers, the findings should not be interpreted as patient-level clinical diagnostic validation. Further studies using subject-wise splitting, external multi-center datasets, 3D volumetric modeling, and multimodal clinical information are required to assess generalizability and potential clinical decision-support applicability.

Background and Objectives: Dental procedure coding is essential for accurate billing, reimbursement, and clinical documentation, yet it remains largely manual, time-consuming, and error-prone. While natural language processing (NLP) has enabled significant advances in automated medical coding, limited work has focused on the dental domain, particularly the assignment of Code on Dental Procedures and Nomenclature (CDT) codes from free-text clinical notes. This study aims to develop and evaluate an artificial intelligence framework that integrates large language models (LLMs) and traditional deep learning methods to automate CDT code extraction from narrative dental documentation. Methods: We evaluated three LLM-based strategies-zero-shot prompting, QLoRA fine-tuning, and parameter-efficient fine-tuning (PEFT) using LoRA-alongside a supervised Bidirectional GRU (Bi-GRU) classifier. Experiments were conducted using a synthetic dataset designed to emulate real-world dental encounters. Structured JSON output schemas, few-shot prompting, and scalable batch inference pipelines were employed to ensure consistent and interpretable predictions. Model performance was assessed using micro- and macro-averaged F1 scores, precision, recall, exact-match accuracy, and Hamming loss. Results: The zero-shot LLM achieved the highest micro-F1 score (0.9614) and perfect recall for frequent CDT codes, demonstrating strong baseline reasoning without task-specific training; however, performance declined for rare procedures and diagnostic code hallucinations were common. Fine-tuning improved domain alignment, with the non-quantized PEFT LoRA model outperforming QLoRA across all metrics, though both fine-tuned LLMs showed tendencies to over-generate plausible but incorrect codes. The Bi-GRU model achieved balanced performance (micro-F1 = 0.9362, macro-F1 = 0.9377) with minimal hallucinations but occasionally missed context-dependent procedures. Conclusions: These findings highlight complementary strengths between LLM-based and supervised approaches. LLMs provide strong contextual understanding and rapid deployment, while traditional models offer stable and precise multi-label classification. This work supports the development of hybrid, schema-constrained systems for scalable dental procedure coding.

Mental disorders such as depression, anxiety, and stress are increasingly prevalent, particularly among young adults. Traditional assessment methods rely on self-reports and resource-intensive clinician interviews, limiting scalability and accessibility. Speech-based machine learning models offer a scalable and non-invasive alternative; however, population-level models often struggle to distinguish disorder-related signals from speaker-specific traits, reducing individual prediction accuracy.

We present SNIPSNP (crisprtools.org/snipsnp), a comprehensive bioinformatics pipeline for designing experiments for CRISPR-induced homology-directed repair (HDR). The tool addresses the critical challenge of Cas9 re-cleavage by simplifying the selection of "blocking" silent variants that are effective at inhibiting RNP binding upon donor-templated editing. SNIPSNP handles complex edits, including indels, and uses multi-objective optimization to balance editing efficiency with biological safety. From user-defined wild-type and desired HDR alleles, the pipeline identifies candidate guides, annotating them with integrated efficiency scores and genome-wide off-target assessments. Uniquely, SNIPSNP evaluates guide binding against the post-edit genome to determine whether the therapeutic variant alone disrupts repeated Cas9 recognition. When necessary, it introduces synonymous blocking variants, prioritizing PAM and seed regions to minimize re-cleavage probability and editing of the wild-type (WT) allele when editing heterozygous variants. All candidate modifications undergo safety profiling and prioritization of known benign variants from dbSNP. We experimentally validated SNIPSNP and benchmarked it on pathogenic inborn error of immunity variants in primary patient T-cells. Across loci, SNIPSNP-designed templates outperform standard "correction-only" strategies, demonstrating enhanced precision editing, and reduced re-cleavage, establishing SNIPSNP as a robust platform for genome editing and disease modeling.

Automated seizure detection from long-term scalp electroencephalography (EEG) remains challenging because seizure windows are sparse, channel configurations vary across patients, and clinically useful systems must maintain strict control of false alarms. This study presents a patient-independent hybrid generative-discriminative framework evaluated on the CHB-MIT cohort under leave-one-patient-out cross-validation. In each outer fold, one patient was reserved exclusively for final testing, whereas channel harmonization, representation learning, feature refinement, model training, and operating-point calibration were performed using only the remaining patients. A fold-specific convolutional β-variational autoencoder ([Formula: see text]) was trained on interictal windows from the training patients only. For each window, latent statistics and reconstruction error were combined with handcrafted seizure-sensitive descriptors, followed by train-only sanitization and supervised feature selection. Candidate boosted-tree classifiers, namely LightGBM, CatBoost, and XGBoost, were trained within the training pool, and operating parameters were selected on validation data only. Final performance was then computed on the held-out patient using window-level metrics with record-wise temporal postprocessing. Across the 24 held-out evaluation cases, corresponding to 23 distinct individuals in the public CHB-MIT release, the framework achieved mean accuracy of 0.9964, mean balanced accuracy of 0.9411, mean specificity of 0.9967, and mean false alarms per hour of 4.7154. Across seizure-positive held-out cases, the mean sensitivity, F1-score, ROC-AUC, and average precision were 0.5505, 0.3543, 0.9496, and 0.2990, respectively. These findings show that leakage-safe hybrid representation learning can support seizure detection while explicitly handling heterogeneous channels, severe class imbalance, and false-alarm control in long-term scalp EEG.