Interpreting machine learning models is facilitated by the widely employed locally interpretable model-agnostic explanation (LIME) technique. However, when extending LIME to signal data, its credibility falters due to perturbation techniques used to generate local datasets. These techniques disrupt temporal dependencies among features, leading to unrealistic data points and potentially misleading explanations. Additionally, LIME faces instability and local fidelity issues, limiting its suitability for real-world applications. The absence of a dedicated LIME package tailored for interpreting signal data further diminishes comprehensibility, especially when applied to models trained on such data. In this paper, we introduce Signal-based LIME (Sig-LIME) to address these limitations. Sig-LIME leverages a novel data generation technique that captures temporal dependence among features, enhancing credibility and stability. It combines a random forest model and heatmaps to provide illuminating explanations for predictions drawn from electrocardiogram (ECG) signals, improving model transparency. Empirical findings underscore the enhanced interpretability and comprehension of model predictions attained by Sig-LIME compared to baseline LIME. Our quantitative evaluation based on an analysis of variance (ANOVA) framework, reveals a notable improvement in stability with Sig-LIME, evidenced by an f-statistic of 0.0 and p-values of 1, indicating a complete absence of variation between multiple runs. Regarding local fidelity, Sig-LIME surpasses the baseline LIME, exhibiting a lower average Euclidean distance of 0.49 compared to 17.24. Sig-LIME excels in generating data more akin to the original, achieving remarkable stability and significantly enhancing credibility and local fidelity in the explanations it generates.

Background: Heart rate variability (HRV) is a key biomarker reflecting autonomic nervous system (ANS) function and neurocardiac regulation. Reduced HRV has been associated with cardiovascular risk, neurodegenerative disorders, and autonomic dysfunction. In Parkinson’s disease (PD), HRV impairments indicate altered autonomic balance, which may be modifiable through structured exercise interventions. This study investigates the effects of aerobic exercise on HRV in patients with PD and evaluates autonomic adaptations to rehabilitation. Methods: A total of 110 patients with PD (55 male, 55 female) participated in a supervised three-month aerobic exercise program. HRV was assessed pre- and post-intervention using electrocardiogram (ECG) recordings. Time-domain and frequency-domain HRV metrics, including standard deviation of RR intervals (SDRR), very-low-frequency (VLF), low-frequency (LF), high-frequency (HF) power, and LF/HF ratio, were analyzed. Principal Component Analysis (PCA) and clustering techniques were applied to identify subgroups of HRV responders based on autonomic adaptation. Results: Significant improvements in HRV were observed post-intervention, with a reduction in LF/HF ratio (p < 0.05), indicating improved autonomic balance. Cluster analysis identified four distinct HRV response subgroups: Strong Responders, Moderate Responders, Mixed/Irregular Responders, and Low Responders. These findings highlight individual variability in autonomic adaptations to exercise. PCA revealed that key HRV parameters contribute differently to autonomic regulation, emphasizing the complexity of HRV changes in PD rehabilitation. Conclusions: This study demonstrates that aerobic exercise induces beneficial autonomic adaptations in PD patients, as reflected by HRV changes. The identification of response subgroups suggests the need for personalized rehabilitation strategies to optimize autonomic function. Further research is warranted to explore the long-term impact of HRV-guided rehabilitation interventions in PD management.

Background and Objectives: Chronic wounds pose a significant healthcare burden due to their prolonged healing times and susceptibility to infection. Electric field (EF)-enabled smart bandages offer a promising solution by combining therapeutic stimulation with real-time physiological monitoring. Materials and Methods: This study assessed a smart bandage integrating spiral stainless steel electrodes delivering a 200 millivolts per millimeter (mV/mm) EF for 5 h daily over 14 days to full-thickness excisional wounds in 100 Sprague–Dawley rats. Vital signs including heart rate (BPM), oxygen saturation (SpO₂), and temperature were monitored continuously. Machine learning models were trained on these data to predict wound healing status. Results: By Day 7, EF-treated wounds demonstrated significantly faster healing, achieving an average wound closure rate of 82.0% ± 2.1% compared to 70.75% ± 2.3% in the control group (p < 0.05). By Day 14, wounds in the experimental group had significantly reduced to 0.01 ± 0.005 cm², while the control group retained a wound size of 0.24 ± 0.03 cm² (p < 0.05). Histological analysis revealed enhanced neovascularization, collagen alignment, and epithelial regeneration in the EF group. Physiological data showed no systemic inflammatory response. Predictive modeling using XGBoost and Random Forest achieved >98% accuracy, with SHAP (SHapley Additive exPlanations) analysis identifying EF exposure and treatment duration as key predictors. Conclusions: The findings demonstrate that EF-based smart bandages significantly enhance wound healing and enable highly accurate prediction of outcomes through machine learning models. This bioelectronic approach holds strong potential for clinical translation.

Background/Objectives: This study investigates the potential of Pulse Transit Time (PTT) derived from Impedance Plethysmography (IPG), Photoplethysmography (PPG), and Electrocardiography (ECG) for non-invasive and cuffless blood pressure monitoring. IPG measures blood volume changes through electrical conductivity, while PPG detects variations in microvascular blood flow, providing essential insights for wearable health monitoring devices. Methods: Data were collected from 100 healthy participants under resting and post-exercise conditions using a custom IPG system synchronized with ECG, PPG, and blood pressure readings to create controlled blood pressure variations. Machine learning models, including Random Forest, Logistic Regression, Support Vector Classifier, and K-Neighbors, were applied to predict blood pressure categories based on PTT and cardiovascular features. Results: Among the various machine learning models evaluated, Random Forest demonstrated effective performance, achieving an overall accuracy of 90%. The model also exhibited robustness, effectively handling the challenge of unbalanced classes, with a 95% confidence interval (CI) for accuracy ranging from 80% to 95%. This indicates its reliability across different data splits despite the class imbalance. Notably, PTT derived from PPG emerged as a critical predictive feature, further enhancing the model’s ability to accurately classify blood pressure categories and solidifying its utility in non-invasive cardiovascular monitoring. Conclusions: The findings affirm the efficacy of using PTT measurements from PPG, IPG, and ECG as reliable predictors for non-invasive blood pressure monitoring. This study substantiates the integration of these techniques into wearable devices, offering a significant advancement for continuous, cuffless, and non-invasive blood pressure assessment.

This study systematically examines pediatric adnexal torsion, proposing a diagnostic approach using machine learning techniques to distinguish it from acute appendicitis. Our retrospective analysis involved 41 female pediatric patients divided into two groups: 21 with adnexal torsion (group 1) and 20 with acute appendicitis (group 2). In group 1, the average age was 10 ± 2.6 years, while in group 2, it was 9.8 ± 21.9 years. Our analysis found no statistically significant age distinctions between these two groups. Despite acute lower abdominal pain being a common factor, group 1 displayed shorter pain duration (28.9 h vs. 46.8 h, p < 0.05), less vomiting (28% vs. 50%, p < 0.05), lower fever incidence (4.7% vs. 50%, p < 0.05), reduced leukocytosis (57% vs. 75%, p < 0.05), and CRP elevation (30% vs. 80%, p < 0.05) compared to group 2. Machine learning techniques, specifically support vector classifiers, were employed using clinical presentation, pain duration, white blood cell counts, and ultrasound findings as features. The classifier consistently demonstrated an average predictive accuracy of 87% to 97% in distinguishing adnexal torsion from appendicitis, as confirmed across various SVM models employing different kernels. Our findings emphasize the capacity of support vector machines (SVMs) and machine learning as a whole to augment diagnostic accuracy when distinguishing between adnexal torsion and acute appendicitis. Nevertheless, it is imperative to validate these results through more extensive investigations and explore alternative machine learning models for a comprehensive understanding of their diagnostic capabilities.

Background/Objectives: This study investigates the application of machine learning (ML) techniques in diagnosing speech sound disorders (SSDs) in Saudi Arabic-speaking children, with a specific focus on phonological biomarkers, particularly Infrequent Variance (InfrVar), to improve diagnostic accuracy. SSDs are a significant concern in pediatric speech pathology, affecting an estimated 10–15% of preschool-aged children worldwide. However, accurate diagnosis remains challenging, especially in linguistically diverse populations. Traditional diagnostic tools, such as the Percentage of Consonants Correct (PCC), often fail to capture subtle phonological variations. This study explores the potential of machine learning models to enhance diagnostic accuracy by incorporating culturally relevant phonological biomarkers like InfrVar, aiming to develop a more effective diagnostic approach for SSDs in Saudi Arabic-speaking children. Methods: Data from 235 Saudi Arabic-speaking children aged 2;6 to 5;11 years were analyzed using several machine learning models: Random Forest, Support Vector Machine (SVM), XGBoost, Logistic Regression, K-Nearest Neighbors, and Naïve Bayes. The dataset was used to classify speech patterns into four categories: Atypical, Typical Development (TD), Articulation, and Delay. Phonological features such as Phonological Variance (PhonVar), InfrVar, and Percentage of Consonants Correct (PCC) were used as key variables. SHapley Additive exPlanations (SHAP) analysis was employed to interpret the contributions of individual features to model predictions. Results: The XGBoost and Random Forest models demonstrated the highest performance, with an accuracy of 91.49% and an AUC of 99.14%. SHAP analysis revealed that articulation patterns and phonological patterns were the most influential features for distinguishing between Atypical and TD categories. The K-Means clustering approach identified four distinct subgroups based on speech development patterns: TD (46.61%), Articulation (25.42%), Atypical (18.64%), and Delay (9.32%). Conclusions: Machine learning models, particularly XGBoost and Random Forest, effectively classified speech development categories in Saudi Arabic-speaking children. This study highlights the importance of incorporating culturally specific phonological biomarkers like InfrVar and PhonVar to improve diagnostic precision for SSDs. These findings lay the groundwork for the development of AI-assisted diagnostic tools tailored to diverse linguistic contexts, enhancing early intervention strategies in pediatric speech pathology.