Search Results (8,313)

Accurate and robust power quality disturbance (PQD) classification is critical for modern electrical grids, particularly in noisy environments. This study presents a comprehensive comparative evaluation of machine learning (ML) and deep learning (DL) models for automatic PQD identification. The models evaluated include Support Vector Machines (SVM), Decision Trees (DT), Random Forest (RF), k-Nearest Neighbors (kNN), Gradient Boosting (GB), and Dense Neural Networks (DNN). For experimentation, a hybrid dataset, comprising both synthetic and real signals, was used to assess model performance. The robustness of the models was evaluated by systematically introducing Gaussian noise across a wide range of Signal-to-Noise Ratios (SNRs). A central objective was to directly benchmark the practical implementation and performance of these models across two widely used platforms: MATLAB R2024a and Python 3.11. Results show that ML models achieve high accuracies, exceeding $95\%$ at an SNR of 10 dB. DL models exhibited remarkable stability, maintaining $97\%$ accuracy for SNRs above 10 dB. However, their performance degraded significantly at lower SNRs, revealing specific confusion patterns. The analysis underscores the importance of multi-domain feature extraction and adaptive preprocessing for achieving resilient PQD classification. This research provides valuable insights and a practical guide for implementing and optimizing robust PQD classification systems in real-world, noisy scenarios. Full article

Background/Objectives: Incomplete spinal cord injury (iSCI) represents a significant challenge in neurorehabilitation, with conventional limitations including recovery plateaus and declining patient motivation. Virtual reality (VR) and augmented reality (AR) have emerged as promising technologies to supplement traditional therapy through gamification and multisensory feedback. This systematic review and meta-analysis evaluates the effectiveness of VR and AR interventions for improving balance and locomotor function in patients with incomplete spinal cord injury. Methods: A systematic review was conducted following PRISMA guidelines, with searches in PubMed, Scopus, Web of Science, Science Direct, and Google Scholar. Randomized controlled trials and high-quality controlled studies evaluating VR/AR interventions in patients with iSCI (American Spinal Injury Association Impairment Scale [AIS] classifications B, C, or D) for a minimum of 3 weeks were included. A random-effects meta-analysis (Standardized Mean Difference, SMD; 95% Confidence Interval, CI) was conducted for the balance outcome. Results: Eight studies were included (n = 142 participants). The meta-analysis for balance (k = 5 studies) revealed a statistically significant improvement with a large effect size (SMD = 1.21, 95% CI: 0.04–2.38, p = 0.046). For locomotor function, a quantitative meta-analysis was not feasible due to a limited number of methodologically homogeneous studies; a qualitative synthesis of this evidence remained inconclusive. Substantial heterogeneity was observed in the balance analysis (I² = 81.5%). No serious adverse events related to VR/AR interventions were reported. Conclusions: VR/AR interventions show potential as an effective adjunctive therapy for improving balance in patients with iSCI, though the benefit should be interpreted with caution due to considerable variability between studies. The current evidence for locomotor function improvements is insufficient to draw conclusions, highlighting a critical need for more focused research. Substantial heterogeneity indicates that effectiveness may vary according to specific intervention characteristics, populations, and methodologies. Larger multicenter studies with standardized protocols are required to establish evidence-based clinical guidelines. Full article

Fuzzy logic has been applied to a wide range of problems, including process control, object recognition, image and signal processing, prediction, classification, decision-making, optimization, and time series analysis. These apply to solar energy systems. Though experts in renewable energy prefer fuzzy logic techniques, their contribution to the decision-making process of solar energy systems lies in the possibility of illustrating risk factors and introducing the concepts of linguistic variables of data from solar energy applications. In solar energy systems, the primary beneficiaries and audience of the fuzzy logic techniques are solar energy policy makers, as it concerns decision-making models, ranking of criteria or weights, and assessment of the potential location of the installation of solar energy plants, depending on the case. In a real-world scenario, fuzzy logic allows easy and efficient controller configuration in a non-linear control system, such as a solar panel. This study attempts to review the role and contribution of fuzzy logic in solar energy based on its applications. The findings from the review revealed that the fuzzy logic application identifies and detects faults in solar energy systems as well as in the optimization of energy output and the location of solar energy plants. In addition, fuzzy model (predicting), hybrid model (simulating performance), and multi-criteria decision-making (MCDM) are components of fuzzy logic techniques. As the review indicated, these are useful as a solution to the challenges of solar energy systems. Importantly, the integration and incorporation of fuzzy logic and neural networks should be recommended for the efficient and effective performance of solar energy systems. Full article

Background: Public discourse on social media plays an increasingly influential role in shaping health-related perceptions and behaviours. Individuals share experiences, concerns, and opinions beyond clinical settings around different issues. X (formerly Twitter) provides a unique lens through which to examine how different treatments are perceived, used, and debated across diverse communities over time. Objective: The study aims to (a) identify the types of ADHD medications mentioned in posts, depending on language and user type; (b) evaluate the popularity of content related to these medications, considering language and user type; (c) analyse temporal changes in the frequency of mentions between 2006 and 2022; and (d) examine the distribution of tweets across different content categories. By addressing these objectives, this study provides insights into public perceptions of ADHD medications, which may help healthcare professionals better understand online discussions and improve their communication with patients, facilitating more informed treatment decisions. Methods: An observational study was conducted analysing 254,952 tweets in Spanish and English about ADHD medications from January 2006 to December 2022. Content analysis combined inductive and deductive approaches to develop a categorisation codebook. BERTWEET and BETO models were used for machine learning classification of English and Spanish tweets, respectively. Descriptive statistical analysis was performed. Results: Overall, stimulant medications were posted more frequently and received higher engagement than non-stimulant medications. Methylphenidate, dextroamphetamine, and atomoxetine were the most commonly mentioned medications, especially by patients, who emerged as the most active users among the English tweets. Regarding medical content, tweets in English contained more than twice the number of mentions of inappropriate use compared to those in Spanish. There was a high content of online medication requests and offers in both languages. Conclusions: In this study, conducted on X, discussions on ADHD medications highlighted concerns about misuse, adherence, and trivialisation, with clear differences between English and Spanish tweets regarding focus and type of user participation. These findings suggest that monitoring social media can provide early signals about emerging trends, helping clinicians address misconceptions during consultations and informing public health strategies aimed at the safer and more responsible use of ADHD medications. Full article

Background/Objectives: Alcohol consumption can increase the risk of dyslipidemia, thereby elevating the risk of cardiovascular diseases. However, the relationship between alcohol consumption patterns and dyslipidemia remains controversial. Based on large-scale cross-sectional data from the Chinese population, this study aims to investigate the correlations between various alcohol consumption behaviors and dyslipidemia among adult residents in China. Methods: Our analysis utilized data from the 2015–2017 China Nutrition and Health Surveillance project, which provides a large, nationally representative sample (N = 52,471). We employed a binary logistic regression model specifically designed for complex sampling frameworks. This model was utilized to assess the relationship between various alcohol consumption behaviors (including daily alcohol intake levels and drinking frequency) and the incidence of hypercholesterolemia, hypertriglyceridemia, low levels of high-density lipoprotein cholesterol (low HDL-C), and elevated levels of low-density lipoprotein cholesterol (high LDL-C). Drinking behaviors were classified into three distinct categories for analysis: China classification (never, moderate, excessive), WHO classification (never, low-risk, medium-risk, high-risk), and drinking frequency (never, <1, 1–3, 4–6, ≥7 times/week). Results: Compared with never drinkers, the risk of hypercholesterolemia was significantly higher in men who were excessive drinkers (aOR = 1.39, 95%CI: 1.24–1.57), medium-risk drinkers (aOR = 1.24, 95%CI 1.01–1.53), high-risk drinkers (aOR = 1.67, 95%CI: 1.4–1.95), and those who drank more than once a week (aOR range: 1.27–1.65), and there was no such association in women (p > 0.05). Compared with never drinkers, the risk of hypertriglyceridemia was higher in male drinkers with excessive drinking (aOR = 1.35, 95%CI: 1.24–1.47), medium-risk drinking (aOR = 1.29, 95%: 1.11–1.50), high-risk drinking (aOR = 1.52, 95%CI: 1.3–1.71), and a drinking frequency more than 1 time/week (aOR range: 1.22–1.38), while in women, it was moderate drinking (aOR = 0.85, 95%CI 0.77–0.94), low-risk drinking (aOR = 0.86, 95%CI 0.78–0.94), and a drinking frequency of more than once a week (aOR = 0.74, 95%CI 0.63–0.87) that reduced the occurrence of hypertriglyceridemia. Compared with non-drinkers, men with any drinking status had a lower risk of low HDL-C (aOR range: 0.38–0.90) and a similar association was also observed in women (aOR range: 0.26–0.84). Compared with never drinkers, male excessive drinkers (aOR = 0.86, 95%CI: 0.77–0.97), medium-risk drinkers (aOR = 0.80, 95%CI:0.65–0.99), high-risk drinkers (aOR = 0.83, 95%CI: 0.70–0.97), and those with a drinking frequency of 1–3 times/week (aOR = 0.89, 95%: 0.79–0.99) had a lower risk of high LDL-C, and there was no such association in women (p > 0.05). Conclusions: Significant gender differences were observed in the effects of alcohol consumption on lipid profiles. Men who were excessive drinkers, medium-risk drinkers, high-risk drinkers, and those who drank more than once a week had a higher risk of hypercholesterolemia and hypertriglyceridemia, but a lower risk of low HDL-C and high LDL-C. In women, moderate drinking was associated with a reduced risk of hypertriglyceridemia. Any alcohol consumption and drinking frequency more than 1 time/week were associated with a lower risk of low HDL-C in women. No significant association was found between alcohol consumption and hypercholesterolemia or high LDL-C in women. Full article

Background and Objectives: Alpha-1 antitrypsin deficiency (AATD) is a rare genetic condition associated with chronic respiratory diseases such as chronic obstructive pulmonary disease (COPD) and emphysema, and with liver involvement through a distinct toxic gain-of-function mechanism. Despite its clinical relevance, AATD remains underdiagnosed and exhibits marked phenotypic heterogeneity. Artificial intelligence (AI) has shown growing potential in respiratory medicine, yet its application to AATD is still limited. This systematic review synthesizes the clinical evidence on AI in AATD, primarily in the respiratory domain and, where available, in hepatic outcomes. Materials and Methods: We conducted a PRISMA-guided search (PubMed, Web of Science, IEEE Xplore) for original, peer-reviewed articles (January 2014–September 2025) applying AI to detection, classification, stratification, or prediction tasks in AATD. Results: Six studies met eligibility criteria. Supervised models (e.g., XGBoost, penalized regression, Transformer-based architectures) and one unsupervised approach were identified. Applications included screening in COPD populations, prediction of emphysema progression from CT, proteomic modeling of lung function, identification of clinical subgroups, and prediction of clinical outcomes in AATD-associated liver disease. External validation and genotype diversity remained limited across studies. Conclusions: Although AI shows promise in improving detection, prognosis, and patient stratification in AATD across both respiratory and hepatic manifestations, the current evidence remains limited. Broader, multicenter validation in genotype-diverse cohorts is required to confirm its clinical utility and support the implementation of precision medicine in AATD. Full article

Objectives: This study aimed to compare the clinical, functional, and radiological outcomes of three different fixation techniques—dorsal locking plate, crossed cortical screw, and a combination of both—used in first metatarsophalangeal (MTP) joint arthrodesis for advanced-stage hallux rigidus. The goal was to provide evidence-based guidance for surgical technique selection. Methods: This retrospective cohort study included 52 patients with advanced hallux rigidus (stage III–IV, Coughlin–Shurnas classification) who underwent surgical treatment between 2023 and 2025 at the Department of Orthopedics and Traumatology of Ankara Etlik City Hospital, with a minimum follow-up of one year. Patients were categorized into three groups according to the fixation technique used. Visual Analog Scale (VAS), American Orthopaedic Foot & Ankle Society (AOFAS) score, and Foot Function Index (FFI) were assessed using validated Turkish-language versions of the questionnaires. Radiological parameters included hallux valgus angle, first toe dorsiflexion angle, distal interphalangeal (DIP) arthritis, and radiographic union—defined as trabecular bridging across at least three cortices on weight-bearing anteroposterior and lateral radiographs. ANCOVA was performed with age as a covariate. Results: A total of 52 patients were included: Group 1 (dorsal plate fixation, n = 19), Group 2 (crossed cortical screw fixation, n = 16), and Group 3 (combined fixation, n = 17). Group 1 patients were significantly older (mean age: 64 ± 6 vs. 55 ± 6 and 59 ± 5 years; p < 0.001). After age adjustment, VAS pain scores were significantly higher in Group 1 compared to Group 3 (mean VAS: 2.8 ± 0.6 vs. 1.9 ± 0.5; p = 0.010). AOFAS scores did not differ significantly (p = 0.166), although Group 2 showed the highest median value (90 [70–93]). FFI scores differed significantly (p < 0.001), with Group 1 reporting worse outcomes (19 [17–31]) than Group 2 (15 [13–22], p = 0.03) and Group 3 (15 [11–16], p = 0.01). Dorsiflexion angle was significantly lower in Group 2 than Group 1 (median 19° vs. 27°; p = 0.04), though all remained within the physiological range. Radiographic union was achieved in 50/52 patients (96.2%), without significant intergroup differences (p = 0.612). Complications included two cases of wound dehiscence in Group 1; no infections, symptomatic non-union, malalignment, or hardware irritation were observed. Conclusions: Crossed cortical screw fixation yielded the most favorable functional outcomes, whereas the combined technique achieved the lowest postoperative pain scores. Dorsal plate fixation alone consistently underperformed. While outcomes were adjusted for age, residual confounding cannot be excluded. These results highlight the importance of tailoring fixation strategy to patient profile, with crossed screw and combined methods representing reliable choices for optimizing postoperative outcomes in advanced hallux rigidus. Full article

Background & Objective: Stroke remains a leading cause of mortality and long-term disability worldwide, demanding rapid and accurate diagnosis to improve patient outcomes. Computed tomography (CT) scans are widely used in emergency settings due to their speed, availability, and cost-effectiveness. This study proposes an artificial intelligence (AI)-based framework for multiclass stroke classification (ischemic, hemorrhagic, and no stroke) using CT scan images from the Ministry of Health of the Republic of Turkey. Methods: We adopted MaxViT, a state-of-the-art Vision Transformer (ViT)-based architecture, as the primary deep learning model for stroke classification. Additional transformer variants, including Vision Transformer (ViT), Transformer-in-Transformer (TNT), and ConvNeXt, were evaluated for comparison. To improve model generalization and handle class imbalance, classical data augmentation techniques were applied. Furthermore, explainable AI (XAI) was integrated using Grad-CAM++ to provide visual insights into model decisions. Results: The MaxViT model with augmentation achieved the highest performance, reaching an accuracy and F1-score of 98.00%, outperforming the baseline Vision Transformer and other evaluated models. Grad-CAM++ visualizations confirmed that the proposed framework effectively identified stroke-related regions, enhancing transparency and clinical trust. Conclusions: This research contributes to the development of a trustworthy AI-assisted diagnostic tool for stroke, facilitating its integration into clinical practice and improving access to timely and optimal stroke diagnosis in emergency departments. Full article

Background/Objectives: Video laryngoscopy is one of the primary methods used by otolaryngologists for detecting and classifying laryngeal lesions. However, the diagnostic process of these images largely relies on clinicians’ visual inspection, which can lead to overlooked small structural changes, delayed diagnosis, and interpretation errors. Methods: AI-based approaches are becoming increasingly critical for accelerating early-stage diagnosis and improving reliability. This study proposes a hybrid Convolutional Neural Network (CNN) architecture that eliminates repetitive and clinically insignificant frames from videos, utilizing only meaningful key frames. Video data from healthy individuals, patients with vocal fold nodules, and those with vocal fold polyps were summarized using three different threshold values with the Structural Similarity Index Measure (SSIM). Results: The resulting key frames were then classified using a hybrid CNN. Experimental findings demonstrate that selecting an appropriate threshold can significantly reduce the model’s memory usage and processing load while maintaining accuracy. In particular, a threshold value of 0.90 provided richer information content thanks to the selection of a wider variety of frames, resulting in the highest success rate. Fine-tuning the last 20 layers of the MobileNetV2 and Xception backbones, combined with the fusion of extracted features, yielded an overall classification accuracy of 98%. Conclusions: The proposed approach provides a mechanism that eliminates unnecessary data and prioritizes only critical information in video-based diagnostic processes, thus helping physicians accelerate diagnostic decisions and reduce memory requirements. Full article

Deploying deep learning (DL) models in real-world environments remains a major challenge, particularly under resource-constrained conditions where achieving both high accuracy and compact architectures is essential. While effective, Conventional pruning methods often suffer from high computational overhead, accuracy degradation, or disruption of the end-to-end training process, limiting their practicality for embedded and real-time applications. We present Dynamic Attention-Guided Pruning (DAGP), a Dynamic Attention-Guided Soft Channel Pruning framework that overcomes these limitations by embedding learnable, differentiable pruning masks directly within convolutional neural networks (CNNs). These masks act as implicit attention mechanisms, adaptively suppressing non-informative channels during training. A progressively scheduled L1 regularization, activated after a warm-up phase, enables gradual sparsity while preserving early learning capacity. Unlike prior methods, DAGP is retraining-free, introduces minimal architectural overhead, and supports optional hard pruning for deployment efficiency. Joint optimization of classification and sparsity objectives ensures stable convergence and task-adaptive channel selection. Experiments on CIFAR-10 (VGG16, ResNet56) and PlantVillage (custom CNN) achieve up to 98.82% FLOPs reduction with accuracy gains over baselines. Real-world validation on an enhanced PlantDoc dataset for agricultural monitoring achieves 60 ms inference with only 2.00 MB RAM on a Raspberry Pi 4, confirming efficiency under field conditions. These results illustrate DAGP’s potential to scale beyond agriculture to diverse edge-intelligent systems requiring lightweight, accurate, and deployable models. Full article

Objectives: Periapical periodontitis, which is a periodontal dysfunction, is a current clinical problem. Due to the frequency of occurrence and the adverse effects they cause, they are considered a social disease. They require detailed diagnostics to implement appropriate treatment. Mathematical calculations based on data obtained from radiological images used in routine clinical practice may help differentiate the forms of periodontitis. This study aimed to evaluate the areas affected by periodontitis in comparison to the healthy tissues of the periapical area. Methods: The study analyzed texture components using the gray-level co-occurrence matrix (GLCM) and the gray-level run-length matrix (GRLM) on an orthopantomogram (OPG) from 50 patients with clinically confirmed periodontitis treated at the Department of Maxillofacial and Plastic Surgery, University of Bialystok. Texture analysis was performed on defined regions of interest (ROIs) to distinguish diseased from healthy tissues. We employed four classification algorithms to assess model performance. Results: The data set included 50 patients, with 76 cases of periodontitis and 50 healthy ROIs. The reference standard was clinical diagnosis confirmed by two specialist doctors. The best-performing algorithm achieved an AUC of 98%. Conclusions: The obtained results showed significant statistical differences in the inflamed regions compared to the control, which may aid in diagnosing and selecting the treatment method for periodontitis. Full article

Background/Objectives: Thrombosis, a critical condition that can have severe consequences, such as myocardial infarction and cerebral infarction, can be induced by steroid drugs. Although the mechanisms for inducing thrombosis are known for some types of steroid drugs, much remains unknown about the differences in the tendency and mechanisms for thrombosis. Methods: To address this knowledge gap, we analyzed the relationship between thrombosis and steroid use by utilizing the U.S. Food and Drug Administration Adverse Event Reporting System database. From the database, we extracted demographic and drug information and information on reported adverse events from 2004 to 2024. We characterized drugs according to physiological function, receptor specificity, and Anatomic Therapeutic Chemical classification and calculated the proportion of steroid drugs that were likely to induce thrombosis. Results: Among steroid drugs, sex hormones such as androgens, progestogens, and estrogens appeared to have particularly high potential for causing thrombotic events. Results of principal component analysis and cluster analysis indicated that sex hormone preparations were associated with an increased risk of venous thrombosis. In addition, cardiovascular medications and mineralocorticoids, which are used to treat diseases of major organs, showed a tendency to induce large-vessel occlusions. Conclusions: These findings may be useful for selecting steroid drugs for patients who are at risk for similar adverse effects. Full article

Background/Objectives: Mosaicism in preimplantation embryos has important implications for embryo selection and reproductive outcomes. This study investigates the age-related frequency of mosaicism, analyzes its subtypes, and evaluates its clinical significance. Methods: A total of 36,506 blastocysts were analyzed using next-generation sequencing-based preimplantation genetic testing for aneuploidy between January 2021 and December 2023. The overall frequencies of euploid, aneuploid, mosaic, and no-call embryos were 20%, 56%, 23%, and 1%, respectively. In this study, we propose a new classification. Embryos were classified into two categories: Mosaic-A, referring to embryos identified as mosaic in standard genetic testing reports, and Mosaic-B, which includes both Mosaic-A and aneuploid embryos containing mosaicism. Results: The proportion of Mosaic-A embryos significantly decreased with maternal age (31% in women <35 years, 30% at 35–37 years, 23% at 38–40 years, 16% at 41–42 years, and 10% in women >42 years). By contrast, Mosaic-B embryos, which include Mosaic-A and aneuploid embryos with mosaicism, increased with age (46%, 49%, 53%, 56%, and 62% across the same age groups). Notably, as maternal age advanced, low-level complex mosaicism decreased, whereas high-level complex mosaicism significantly increased (p < 0.001, chi-square test for trend). Other mosaicism subtypes followed similar trends. These findings suggest that the increase in high-level complex mosaicism resulted from a higher incidence of post-zygotic mitotic errors occurring earlier in development and affecting a larger proportion of cells in older women. Conclusions: This study underscores the significance of incorporating a broader classification of mosaicism, including Mosaic-A and B, to enhance understanding of the biological behavior of mosaic embryos according to age and highlights the clinical importance of cryopreserving high-level or complex mosaic embryos for transfer in women of advanced age. Full article

Assessing teaching behavior is essential for improving instructional quality, particularly in Physical Education, where classroom interactions are fast-paced and complex. Traditional evaluation methods such as questionnaires, expert observations, and manual discourse analysis are often limited by subjectivity, high labor costs, and poor scalability. These challenges underscore the need for automated, objective tools to support pedagogical assessment. This study explores and compares the use of Transformer-based language models for the automatic classification of teaching behaviors from real classroom transcriptions. A dataset of over 1300 utterances was compiled and annotated according to the teaching styles proposed in the circumplex approach (Autonomy Support, Structure, Control, and Chaos), along with an additional category for messages in which no style could be identified (Unidentified Style). To address class imbalance and enhance linguistic variability, data augmentation techniques were applied. Eight pretrained BERT-based Transformer architectures were evaluated, including several pretraining strategies and architectural structures. BETO achieved the highest performance, with an accuracy of 0.78, a macro-averaged F1-score of 0.72, and a weighted F1-score of 0.77. It showed strength in identifying challenging utterances labeled as Chaos and Autonomy Support. Furthermore, other BERT-based models purely trained with a Spanish text corpus like DistilBERT also present competitive performance, achieving accuracy metrics over 0.73 and and F1-score of 0.68. These results demonstrate the potential of leveraging Transformer-based models for objective and scalable teacher behavior classification. The findings support the feasibility of leveraging pretrained language models to develop scalable, AI-driven systems for classroom behavior classification and pedagogical feedback. Full article