Background: Diabetes mellitus affects 463 million people worldwide and necessitates continuous blood glucose monitoring. Current glucose prediction systems often lack efficiency, and real-time prediction is essential for timely clinical intervention. Objectives: This study aims to develop and validate a novel Convolutional Recurrent Neural Network (CRNN) enhanced with bio-inspired algorithms to improve blood glucose prediction and enable real-time detection of hypoglycemia and hyperglycemia. Materials and methods: The proposed framework employs a CRNN architecture that combines Convolutional Neural Networks (CNNs) for feature extraction with Long Short-Term Memory (LSTM) layers for temporal sequence learning. The model was trained and evaluated using the HUPA-UCM diabetes dataset. Additionally, the study benchmarks the proposed model against 19 traditional Machine Learning (ML) algorithms and compares it with state-of-the-art methods from the literature. Results: The proposed approach demonstrates superior predictive capability, consistently delivering promising results across multiple evaluation frameworks. The model achieves clinically acceptable prediction intervals, confirming its effectiveness in enhancing the accuracy and reliability of blood glucose prediction for diabetes management. Conclusion: The findings demonstrate that the proposed CRNN model, enhanced with bio-inspired algorithms, provides an effective and reliable solution for real-time blood glucose prediction. By outperforming conventional ML methods and achieving clinically acceptable accuracy levels, the model shows strong potential for integration into intelligent diabetes management systems to support timely clinical decisions and improve patient outcomes.

Vision foundation model for 3D magnetic resonance imaging segmentation, classification, and registration.

Multilabel classification tackles scenarios where each sample concurrently belongs to multiple binary classes, referred to as labels. The task of learning multilabel datasets is harder than single-label classification. In addition to its inherent complexity, low-density label datasets, noisy labels, and complex relationships between labels make this problem extremely difficult. In this paper, we present a new approach, false flag labeling , which is based on modifying the labels of the instances of the training set to obtain a new training set that can improve the classification model’s performance. We assume that adding new relevant labels to the training set or removing relevant labels from it may improve the results of the learning algorithm, making the search for separation surfaces easier. We do not assume that the added labels correspond to the actual labels of the instances that were missed in the dataset labeling process, nor do we assume that the removed labels correspond to erroneously labeled instances. We use an evolutionary algorithm to approach the false labeling process as an optimization process. An extensive comparison using 50 datasets and seven classification models demonstrates the advantageous performance of our approach. • We present a new method for improving multi-label classification. • The method relies on modifying the training set’s labels to improve results. • The comparison uses four different metrics. • We study the method’s performance in the context of missing labels.

Machine learning to predict complications after salvage surgery in head and neck cancers.

Integrating Reproductive and Clinical Variables to Predict Postpartum Disability Outcomes in Multiple Sclerosis Using Machine Learning.

Manual cacao pod harvesting involves intricate tasks that significantly impact the overall efficiency of the process. This study comprehensively analyses motion and time allocation during manual cacao pod harvesting. Employing both visual-based and motion-capture methodologies, this research offers insights into task categorization and time utilization. Visual-based analysis identified four primary tasks—searching for, severing from trees, lifting and carrying — revealing that searching consumed the most time (49%), followed by severing (33.2%) and lifting and carrying (17.8%). The lifting and carrying task was further divided into two sub-tasks: lifting pods and transporting them to a designated collection point. Models were trained to classify tasks using the Support Vector Machine with Radial Basis Function (SVM RBF) algorithm in motion capture. However, discrepancies were observed between observed and predicted time usage for tasks, potentially due to nuances in motion execution among workers. Performance metrics highlighted limitations in task identification by the classification models. Suggestions provided for model improvement included incorporating diverse motion patterns in the training data and considering sub-task classifications for better accuracy. This study underscores and highlights subtasks during the harvest that can be easily automated to quickly enhance harvest process efficiency. • Developed the first video–motion analysis of manual cacao harvesting tasks. • Established standardised hierarchical task and sub-task taxonomy. • Analysis revealed that search and severing are the dominant efficiency bottlenecks. • Validated a ML model using joint-angle kinematics to classify harvesting tasks. • Identified tasks suitable for mechanisation and guide design for robotics. Science4Impact Statement (S4IS)This study investigates manual cacao pod harvesting, revealing that significant time is consumed in searching for ripe pods (49%) and severing them from trees (33.2%). These findings suggest potential automation opportunities that could streamline harvesting processes. The results support regulatory frameworks by demonstrating how improved harvesting methods can enhance food security and labour efficiency. Implementing these advancements could contribute to sustainable agricultural practices, ultimately addressing climate change and food production challenges, thereby aligning with societal goals for agricultural innovation.

Robust detection of soluble solids content in winter jujube using contrastive learning and spectroscopy: Mitigating the impact of measurement position

Developing fair classification models is a crucial aspect of machine learning research. However, unintended distortion in training data - biased data - can lead to discriminatory decisions. In this paper, we developed a workflow for detecting and mitigating bias in data using a shallow, interpretable machine learning models: the Generalized Matrix Learning Vector Quantization. We extent the approach by a relevance-based analysis to identify and reduce bias in the data. Combining similarity metric adaptation and relevance-based analysis, we can develop fair classification models that minimize the influence of bias in the data. Our results demonstrate that this method is effective in reducing bias in classification models and therefore supports fair decision-making. • Model native derived solution for bias mitigation avoiding surrogate methods. • Workflow to examine a bias hypothesis and, if applicable, mitigate the bias in a transparent manner. • Mitigated bias model performance is evaluated for significant differences to the original models performance.

Leveraging knowledge distillation for lightweight and interpretable deep learning in Ethiopian medicinal plant classification

Purpose To evaluate large language model (LLM)-based strategy performance for extraction and classification of incidental findings from whole-body (WB) imaging reports, particularly strategies incorporating Oncologically Relevant Findings Reporting and Data System (ONCO-RADS). Materials and Methods In this retrospective bicenter study, authors included all WB MRI reports from January 2016 to December 2023 at a referral center (internal dataset). Two observers extracted all incidental findings, and patient records were used to confirm final diagnoses. First, authors evaluated ONCO-RADS performance and the reproducibility of its incidental finding classifications by six radiologists. Then, authors evaluated the accuracy of three LLM-based strategies: (a) a fine-tuned DeBERTa/medical named entity recognition (NER) model; (b) zero-shot LLMs (ChatGPT-o1 [OpenAI], Gemini-2.5-Pro [Google]); and (c) reference-guided prompting of these LLMs using ONCO-RADS. Authors then expanded these strategies to an external dataset of 605 reports with multiple imaging techniques (405 WB MRI; 100 fluorodeoxyglucose PET/CT; and 100 chest-abdomen-pelvis CT acquisitions) from January 2022 to January 2025. Results The internal dataset included 823 patients (mean age, 63.7 years ± 11.7 [SD]; 457 male patients) with 1488 WB MRI reports. The average interobserver reproducibility of ONCO-RADS incidental finding classifications was excellent (Cohen κ, 0.87). The per-report accuracies of ONCO-RADS-guided LLMs (95.6% [151 of 158] and 86.7% [137 of 158] for ChatGPT-o1 and Gemini-2.5-Pro, respectively) were higher than those of the medical NER (69.0% [109 of 158]) and zero-shot LLMs (57.0% [90 of 158] and 70.9% [112 of 158] for ChatGPT-o1 and Gemini-2.5-Pro, respectively) (P < .001). In the external test set (mean age, 60.6 years ± 12.9; 330 male patients), the per-report accuracies of ONCO-RADS-guided ChatGPT-o1 (83.5% [505 of 605]) and Gemini-2.5-Pro (82.0% [496 of 605]) were higher than those of the models without ONCO-RADS prompting (63.1% [382 of 605] and 61.2% [370 of 605], respectively) and the medical NER (55.7% [337 of 605]) (P < .001). Conclusion Reference-guided prompting of the LLMs ChatGPT-o1 and Gemini-2.5-Pro with ONCO-RADS improved their performance in extracting and classifying incidental findings on WB imaging reports compared with zero-shot prompting and medical NER. Keywords: Large Language Models, Incidental Findings, Whole-Body MRI Supplemental material is available for this article. © RSNA, 2026.

Comprehensive chemical fingerprinting by LC×LC-fluorescence and data-driven chemometric modelling for unsupervised classification.

Mango leaf disease represents a significant threat to fruit quality and yield, necessitating highly accurate, real-time detection systems. However, existing Deep Learning approaches, particularly Transformer-based models, often suffer from prohibitive computational complexity (quadratic scaling), limiting their deployment on resource-constrained edge devices. To address this challenge, this study introduces MangoMamba, a novel lightweight hybrid architecture specifically optimized for mobile deployment. The proposed model integrates Multi-Scale Mamba Mixers with Large-Kernel Attention mechanisms within a hierarchical four-stage framework, enabling linear computational complexity while preserving global receptive fields. Experimental evaluations were conducted on the MangoLeafBD dataset and the newly curated VN-MangoLeaf dataset, which comprises 7000 images of Vietnamese mango varieties. Results demonstrate that MangoMamba achieves competitive classification accuracies of 99.75% and 98.71% on the respective datasets. Crucially, the model exhibits exceptional efficiency with only 5.8 million parameters and an inference latency of 1.46 ms per image on T4 GPU, approximately 80 times faster than recent ViX-MangoEFormer architectures. Furthermore, the practical feasibility of the proposed approach is validated through a functional Android application capable of offline inference (100–300 ms latency) on standard smartphones. These findings confirm that MangoMamba establishes a new competitive trade-off between accuracy and efficiency for smart agriculture applications. • MangoMamba achieves ≤ 1.52 ms GPU inference with competitive accuracy. • Hybrid Mamba-Attention model offers linear complexity for disease detection. • VN-MangoLeaf dataset introduced: 7000 images from Vietnam with Red Rust. • Three-phase curriculum learning enables cross-geographical knowledge transfer. • Lightweight ≤ 25 MB model enables 100-300 ms smartphone-based detection.

A cost-efficient deep learning model for audio classification prioritized by application needs

CEA-Net: A multi-modal model for corn disease classification with dynamic fusion and cross-layer connection mechanism

CSBAM-MobileNet: a lightweight attention-enhanced deep learning model for student attentiveness classification

A navigation-guided 3D breast ultrasound scanning and reconstruction system for automated multi-lesion spatial localization and diagnosis.

The primary load-bearing structures of suspension bridges are the main cables, which are constructed with high-tensile-strength steel wires. Throughout the service life of a suspension bridge, the main cables not only endure cyclic loading from various loading sources but also from severe environmental conditions. These long-term applied loading conditions may result in significant deterioration of material characteristics and potentially cable failure, compromising both the longevity and security of the suspension bridge. Thus, analyzing corrosion patterns based on the main cables’ high-tensile-strength steel wires and evaluating their corrosion intensity are critically important for civil engineers. This paper utilizes a copper-accelerated salt spray test to fast generate samples of four distinct corrosion stages of steel wires. By employing the semantic segmentation model Deeplabv3+, the corrosion positions can be determined. By utilizing three image classification models—ResNet50, ShuffleNet, and DFL (Discriminative Filter Bank Learning), the stages of corrosion in samples were classified and analyzed as a reference for engineering applications.

Free water in the hippocampal cingulum as a Radiomic biomarker for Identifying inflammatory neuropsychiatric Lupus: A cross-sectional case-control study.

Toward early warning of unsafe behavior of excavator operators under time pressure: experimental evidence and EEG-based detection via RCF-IncepLite model.

Explainable machine learning framework for the molecular classification of triple negative breast cancer.