Novel velocity model for the quantitative characterization of pleural sliding, in vivo multicenter clinical study on RF lung ultrasound data.

Data-driven classification in optical correlation systems

A Machine learning framework for calf muscle fatigue assessment using IMU sensors and EMG-Based labeling.

Boosted sliced regression for dimension reduction in binary classification

Integration of high-throughput fluorescence and UV–vis spectroscopy for goat milk adulteration detection: comparison of one-class, binary classification and regression models

Reimagining dry eye disease management - a multimodal approach targeting the key pathophysiological drivers.

High-throughput image-based pupal sex classification in Aedes aegypti using convolutional neural network models for sterile insect technique applications.

Anaerobic digestion (AD) is often highlighted as a potential solution for waste management, sanitation, renewable energy production, and nutrient recovery in resource-limited settings. However, post-implementation performance of biogas systems in sub-Saharan Africa remains poorly documented., with many AD systems failing to achieve optimal outcomes. This study evaluates 61 small- and medium-scale biogas systems in Southern Malawi using a qualitative, on-site assessment combined with interviews, participatory observation, and socio-technical analysis. We develop and apply a holistic evaluation framework that integrates engineering design, system functionality, value creation, and user experience, moving beyond the commonly reported binary classifications of success and failure. None of the systems achieved what we term “blue-ribbon success,” defined as which represents an ideal scenario where a system flawlessly delivers all anticipated benefits. Instead, a small number approached “contextual success,” where systems maximised benefits, minimised negative environmental impacts while adapting to local constraints and user realities. Across the spectrum of outcomes, underperformance was most strongly associated with poor design and implementation practices, underpinned by fragmented knowledge and weak coordination among developers, installers, users, and policymakers. These gaps encompass issues around training, maintenance, and operation, ultimately affecting the success and long-term sustainability of biogas systems. Addressing these deficiencies through improved design practices, targeted education, and sustained technical support is essential to improving the viability and value of AD systems where they are needed, suitable and feasible. In fact, ascertaining where they are needed, suitable, and feasible may be the most important step during project and engineering design.

To review recent advances in the use of artificial intelligence (AI) to address shortcomings in assessing and improving surgical performance/training by automating surgical skills assessment and feedback. We searched PubMed for studies published between 2015 and 2025 pertaining to AI for surgical training. Search terms included 'artificial intelligence or 'machine learning' or 'deep learning' and 'surgical feedback' or 'surgical training' or 'surgical skill'. Articles were identified with special attention given to those published in the last 5 years with a focus on AI for surgical skill assessment or feedback. Artificial intelligence has been used to successfully automate surgical skill assessment across a variety of surgical disciplines via approaches such as kinematics, sabermetrics, computer vision, and gesture analysis. Many of these studies have developed AI models capable of a binary classification of skill (novice vs expert), which demonstrate concordance when verified against ground truths from human raters. Based on these skills assessments, AI approaches may be further leveraged to generate automatic feedback, which has proven effective in improving surgeon performance metrics, particularly for underperformers. AI has also shown utility in categorising and analysing the content and impact of live surgical feedback, enabling more efficient analysis of how feedback can be best delivered to trainees. Artificial intelligence is a promising tool for augmenting surgical training and improving the objectivity and scalability of surgical skill assessment and feedback. To date, AI models are adept at detecting relatively large differences in surgical performance and providing rudimentary feedback. Further work is required to create models capable of doing more fine-tuned skill assessments and generating more detailed, constructive feedback.

The Yin and Yang of tertiary lymphoid structures in primary liver cancer.

A unified framework for EEG seizure detection using universum-integrated generalized eigenvalues proximal support vector machine.

Precise and high-throughput origin discrimination for green coffee beans by mass spectrometry-based metabolic analysis.

The development of a robust deep learning architecture that is not easily affected by overfitting is an important factor in improving the performance of medical image classification systems. This study aims to assess the ability of the DenseNet121 architecture to classify histopathological images into two categories. The model utilizes pre-trained weights from ImageNet and is adjusted through fine-tuning, while geometric data augmentation techniques are performed to increase sample variation. The training process utilizes the AdamW optimizer and the Binary Cross-Entropy loss function, with performance assessment using binary classification metrics. The test results show that DenseNet121 achieved a training accuracy of 98.96%, a validation accuracy of 97.72%, and a testing accuracy of 97.09%, indicating consistent performance at each stage and no signs of overfitting. This finding indicates that DenseNet121 has great potential as an effective structure in histopathological image classification systems.

Chronic wound image classification is important for supporting the assessment of conditions such as diabetic foot ulcers (DFU) and pressure ulcers (PU). While convolutional neural network (CNN)–based approaches have shown promising results, most previous studies focus on binary classification and rarely evaluate robustness in multiclass chronic wound scenarios. This study investigates multiclass classification of chronic wound images, distinguishing DFU, PU, and Normal Skin, using ResNet-50 and ResNeXt-50 architectures. A total of 2,146 publicly available images were stratified at the image level into training (70%), validation (15%), and test (15%) sets. Both models were trained under an identical configuration using data augmentation and class-weighted loss. On clean test images, ResNet-50 and ResNeXt-50 achieved strong and comparable performance, with accuracies of 0.9877 and 0.9938 and macro-averaged F1-scores of 0.9866 and 0.9928, respectively. Robustness was evaluated by applying Gaussian blur at the inference stage to simulate image defocus. Under stronger blur (σ = 2.0), ResNeXt-50 maintained higher performance (accuracy 0.9723, macro-F1 0.9679) than ResNet-50 (accuracy 0.9200, macro-F1 0.9123). These results highlight the contribution of this study in evaluating robustness to blur in multiclass chronic wound image classification, while emphasizing that robustness is limited to resistance against image blur or defocus.

The long-term physiologic effects of thyroid problems make them one of the most important endocrine disorders. Even if a lot of machine learning and deep learning techniques have been presented out for the early detection of thyroid disease, it is still difficult to achieve reliable and clinically accurate multi-class diagnostic performance. In this work, we suggest an Enhanced Extreme Learning Machine (EELM) that uses Drop-Connect regularization to enhance generalization and reduce over-fitting that is frequently seen in traditional ELM models. The pipeline for the suggested framework consists of seven steps: data preprocessing, model building, training, and evaluation. To simulate a clinically relevant diagnostic scenario, the model was assessed on a unified four-class thyroid classification task (hypothyroidism, hyperthyroidism, sick-euthyroid, and normal). The suggested EELM demonstrated steady and reliable multi-class performance with an average accuracy of approximately 82% under 10-fold cross-validation. The model achieved up to 99.89% accuracy in comparative binary classification studies (e.g., hypothyroid vs. normal), indicating the better division of some thyroid diseases. Accuracy, precision, recall, specificity, sensitivity, F1-score, ROC, and AUC measures were used to evaluate performance. The suggested method's robustness and significance were validated statistically using ANOVA and paired t-tests. Significant improvements over baseline models were confirmed by statistical validation with paired t-tests and ANOVA (p < 0.05). Overall, the findings show that the suggested EELM offers a clinically applicable, statistically supported, and computationally effective method for classifying thyroid diseases.

Rapid and sensitive detection of plant pathogens, such as the Avocado Sunblotch Viroid (ASBVd), is essential for early disease management and agricultural biosecurity. Yet, most current diagnostic methods not only require relatively large sample inputs but also often lack the ultrasensitivity required for reliable detection with scarce or minimally collected plant material. Here, we report a novel low-input but ultrasensitive diagnostic platform that integrates isothermal recombinase polymerase amplification (RPA), CRISPR-Cas12a detection, and a solid-state nanopore array for the detection of ASBVd. The system leverages CRISPR-Cas12a collateral cleavage activity to generate single-bead fluorescent signals, which are captured by a nanopore array through pressure-driven blockage. Our platform achieves a detection limit down to 1.68 copies/μL while using only 40 nL of bead-fluorophore mixture per readout, which is over 100-fold less than conventional assays based on fluorescent readout using an imaging reader, enabling detection from minimal avocado sample collection. We demonstrate robust binary classification of ASBVd-positive and -negative samples from multiple avocado tissue types and orchards in California. The assay requires just 60 min and operates entirely under isothermal conditions, avoiding the need for bulky PCR instruments and supporting on-site deployment with minimal equipment. This method provides a promising platform for field-deployable, ultrasensitive, and low-input diagnostics of viroids and other low-titer pathogens in plant or clinical settings.

Parkinson's disease (PD) and chronic obstructive pulmonary disease (COPD) are prevalent conditions with substantial impact on quality of life and health care systems. Both disorders affect voice production through different physiological mechanisms, yet neither condition has a widely adopted objective biomarker for routine clinical use. Voice analysis has emerged as a non-invasive digital biomarker candidate, but existing studies have largely focused on binary classification within a single disorder or language. This study aimed to evaluate whether an unified multiclass machine learning (ML) framework applied to sustained vowel "a" phonation can discriminate between PD, COPD, and healthy controls (HC) across linguistically distinct cohorts. Sustained vowel recordings were analyzed from Swedish speaking individuals with COPD and HC, and English-speaking individuals with PD and HC, collected under comparable mobile recording conditions. Acoustic features included baseline voice measures and Mel Frequency Cepstral Coefficients. A soft voting ML framework integrating support vector machine, random forest, CatBoost, and light gradient boosting classifiers was trained using nested cross validation with hyperparameter optimization. Data were partitioned at the participant level into a development cohort and an independent test cohort. Model performance was evaluated using accuracy, macro averaged precision, recall, F1 score, receiver operating characteristic analysis, and confusion matrices. Model interpretability was assessed using Shapley additive explanations and vowel space analysis. The final soft voting classifier achieved robust multiclass discrimination on the participant disjoint independent test set, with an overall accuracy of 0.842 and a macro averaged F1 score of 0.839. Classification performance differed across groups, with the highest performance observed for PD, intermediate performance for HC, and lower performance for COPD. Misclassifications occurred primarily between HC and COPD, while confusion between PD and COPD was minimal. Feature attribution analysis revealed class dependent relevance patterns, and vowel space analysis demonstrated subtle but consistent group level differences. These findings demonstrate the feasibility of using an explainable soft voting machine learning framework applied to sustained vowel phonation to distinguish between neurologically and respiratory driven voice impairments across linguistic contexts. The study supports voice as a promising digital biomarker modality for multiclass clinical discrimination using mobile recordings.

The high rate of Internet of Things (IoT) ecosystem development presents a significant security problem because of the heterogeneity of devices, dynamic network behaviour, and the lack of built-in security features. Traditional intrusion detection systems (IDS) cannot detect multi-stage attacks to IoT systems, and this issue is mainly due to their use of single-modality information and the poor temporal modeling properties of these applications. This paper seeks to overcome these shortcomings by offering the explainable multimodal temporal deep learning framework of intrusion detection (EMT-IDNet). The proposed model combines different data sources network traffic, logs on host operating system and IoT telemetry using modality-specific encoders and cross-modal attention-based fusion block. Temporal attention module gets ultimate attack sequences and detects key time moments related to malicious actions. Additionally, an attention-based hierarchical multi-layered perceptron (TA-MLP) classifier can be used to identify binary and multi-class intrusion with high efficiency and interpretability as demonstrated by visualization of attention weights. Empirical testing on TON-IoT data reveals that EMT-IDNet has a binary classification with 99.98% accuracy, 99.99% precision, 99.97% recall and 99.98% F1-score, and its AUC value is 0.9999. The framework offers significant temporal as well as modality-level explanations, which increase trust and practical utility to applications in the real world IoT security.

Optical coherence tomography (OCT) is extremely useful in the screening and detection of oral cancers. But various challenges, such as its subjectivity, operator dependence in interpretation, and lack of quantitative outcomes, have been delaying its adoption in clinical decision-making. The objective of this research is to quantify various A-scan features embedded in the OCT signal through advanced signal processing techniques and machine learning algorithms.Ex vivoimaging of biopsied oral tissues (normal mucosa, carcinomain situ(CIS), well-differentiated, and poorly differentiated oral squamous cell carcinoma) was performed using a spectral domain OCT system. Our A-scan dataset consisted of representative 1D signals obtained from different regions of the tissue bed. A set of 12 time- and 8 frequency-domain features were computed on each A-scan, and ten machine learning models were evaluated for binary and multi-label classification. LightGBM achieved the highest performance in both binary (accuracy: 0.8847,F1: 0.8878, AUC: 0.9539) and multi-label classification (accuracy: 0.8248,F1: 0.8201, AUC: 0.964). LightGBM was selected as the final model based on its superior and consistent performance across both classification paradigms. Our proof-of-concept feasibility study demonstrates good accuracy for differentiating oral mucosal tissues, highlighting the biomarker signature of A-scans.

Implant placement in the posterior maxilla is often complicated by anatomical limitations such as reduced alveolar bone volume, sinus pneumatization, sinus septa, and variations in ridge morphology. Accurate radiographic evaluation of these structures is therefore essential for predictable implant planning in the maxillary first molar region (MFMR). The aim of this study was to investigate the relationships between alveolar bone height (ABH), bucco-palatal crest thickness (CT), buccal concavity, sinus septa (SS), sinus pathology (SP), and demographic factors (age and gender), as well as jaw side and tooth presence in the MFMR, and to evaluate their relevance for implant planning. A total of 798 cone-beam computed tomography (CBCT) images from 399 individuals were retrospectively analyzed. ABH and CT were measured from the alveolar crest to a point 2mm coronal to the maxillary sinus membrane. The presence of concavity, SS, and SP was recorded using a binary (present/absent) classification. ABH and CT were significantly greater in dentate sites compared to edentulous sites (p < 0.05). Age was significantly associated with CT, buccal concavity, SS, and SP (p < 0.05). Tooth presence was significantly associated with concavity and SP, gender with SP, and jaw side with SS (p < 0.05). Anatomical variations in the MFMR are significantly influenced by demographic and site-related factors. Comprehensive CBCT-based evaluation of alveolar bone dimensions and sinus-related structures is essential for individualized and risk-aware implant planning in this region.