Class imbalance is a prevalent challenge in machine learning that arises from skewed data distributions in one class over another, causing models to prioritize the majority class and underperform on the minority classes. This bias can significantly undermine accurate predictions in real-world scenarios, highlighting the importance of the robust handling of imbalanced data for dependable results. This study examines one such scenario of real-time monitoring systems for fall risk assessment in bedridden patients where class imbalance may compromise the effectiveness of machine learning. It compares the effectiveness of two resampling techniques, the Synthetic Minority Oversampling Technique (SMOTE) and SMOTE combined with Edited Nearest Neighbors (SMOTEENN), in mitigating class imbalance and improving predictive performance. Using a controlled sampling strategy across various instance levels, the performance of both methods in conjunction with decision tree regression, gradient boosting regression, and Bayesian regression models was evaluated. The results indicate that SMOTEENN consistently outperforms SMOTE in terms of accuracy and mean squared error across all sample sizes and models. SMOTEENN also demonstrates healthier learning curves, suggesting improved generalization capabilities, particularly for a sampling strategy with a given number of instances. Furthermore, cross-validation analysis reveals that SMOTEENN achieves higher mean accuracy and lower standard deviation compared to SMOTE, indicating more stable and reliable performance. These findings suggest that SMOTEENN is a more effective technique for handling class imbalance, potentially contributing to the development of more accurate and generalizable predictive models in various applications.

Glaucoma is a progressive optic nerve disease and a leading cause of irreversible blindness worldwide. Early and accurate detection is critical to prevent vision loss, yet traditional diagnostic methods such as optical coherence tomography and visual field tests face challenges in accessibility, cost, and consistency, especially in under-resourced areas. This study evaluates the clinical applicability and robustness of three machine learning models for automated glaucoma detection: a convolutional neural network, a deep neural network, and an automated ensemble approach. The models were trained and validated on retinal fundus images and tested on an independent dataset to assess their ability to generalize across different patient populations. Data preprocessing included resizing, normalization, and feature extraction to ensure consistency. Among the models, the deep neural network demonstrated the highest generalizability with stable performance across datasets, while the convolutional neural network showed moderate but consistent results. The ensemble model exhibited overfitting, which limited its practical use. These findings highlight the importance of proper evaluation frameworks, including external validation, to ensure the reliability of artificial intelligence tools for clinical use. The study provides insights into the development of scalable, effective diagnostic solutions that align with regulatory guidelines, addressing the critical need for accessible glaucoma detection tools in diverse healthcare settings.

Background: This study evaluated the performance of machine learning models trained on two different datasets of knee X-ray images annotated with Kellgren–Lawrence grades. Methods: Learning curves indicated that one model experienced poor training, characterized by underfitting, while the other model demonstrated effective training with proper convergence. The poorly trained model appeared to perform adequately on its internal test set but failed to generalize to an external dataset, yielding suboptimal results. Results: In contrast, the well-trained model not only performed well on its internal validation but also showed adequate performance when tested on the external dataset. Conclusions: These findings highlight the importance of examining learning curves to assess model training quality and the critical necessity of external testing to evaluate generalizability. Most existing studies lack external validation, raising concerns about the reliability of their reported performance. This study emphasizes that without external testing, models may not perform as expected in real-world clinical settings, potentially impacting clinical decision-making for surgical interventions. The results advocate for the inclusion of external validation in model evaluation and the assessment of model convergence using learning curves to ensure the development of robust and generalizable tools for knee osteoarthritis severity assessment and other applications.

Background: High in-domain accuracy in healthcare machine learning (ML) models does not guarantee reliable clinical performance, especially when training and validation protocols are insufficiently robust. This paper presents a standardized framework for training and validating ML models intended for classifying medical conditions, emphasizing the need for clinically relevant evaluation metrics and external validation. Methods: We apply this framework to a case study in knee osteoarthritis grading, demonstrating how overfitting, data leakage, and inadequate validation can lead to deceptively high accuracy that fails to translate into clinical reliability. In addition to conventional metrics, we introduce composite clinical measures that better capture real-world utility. Results: Our findings show that models with strong in-domain performance may underperform on external datasets, and that composite metrics provide a more nuanced assessment of clinical applicability. Conclusions: Standardized training and validation protocols, together with clinically oriented evaluation, are essential for developing ML models that are both statistically robust and clinically reliable across a range of medical classification tasks.

This study presents a clinical utility-driven machine learning framework for retinal Optical Coherence Tomography classification, addressing challenges posed by manual interpretation variability and dataset heterogeneity. The methodology integrates biomimetic data partitioning, deep biomarker extraction via pretrained VGG16 networks, and automated model selection optimized for clinical decision-making. Stratified data curation preserved pathological distributions across training, validation, and testing subsets, while SMOTE optimization mitigated class imbalance. Cross-pathology testing evaluated generalizability on anatomically distinct retinal conditions excluded from training, assessing the framework’s robustness to unseen pathologies. Clinical utility metrics prioritized alignment with ophthalmological imperatives, emphasizing negative predictive value to minimize false negatives and enhance diagnostic reliability. The framework advances AI-driven Optical Coherence Tomography diagnostics by harmonizing computational performance with patient-centered outcomes, enabling standardized disease detection across diverse clinical datasets through robust feature generalization.

Background: This study investigates the application of machine learning models to classify electrocardiogram signals, addressing challenges such as class imbalances and inter-class overlap. In this study, “normal” and “abnormal” refer to electrocardiogram findings that either align with or deviate from a standard electrocardiogram, warranting further evaluation. “Borderline” indicates an electrocardiogram that requires additional assessment to distinguish benign variations from pathology. Methods: A hierarchical framework reformulated the multi-class problem into two binary classification tasks—distinguishing “Abnormal” from “Non-Abnormal” and “Normal” from “Non-Normal”—to enhance performance and interpretability. Convolutional neural networks, deep neural networks, and tree-based models, including Gradient Boosting Classifier and Random Forest, were trained and evaluated using standard metrics (accuracy, precision, recall, and F1 score) and learning curve convergence analysis. Results: Results showed that convolutional neural networks achieved the best balance between generalization and performance, effectively adapting to unseen data and variations without overfitting. They exhibit strong convergence and robust feature importance rankings, with ventricular rate, QRS duration, and P-R interval identified as key predictors. Tree-based models, despite their high performance metrics, demonstrated poor convergence, raising concerns about their reliability on unseen data. Deep neural networks achieved high sensitivity but suffered from overfitting, limiting their generalizability. Conclusions: The hierarchical binary classification approach demonstrated clinical relevance, enabling nuanced diagnostic insights. Furthermore, the study emphasizes the critical role of learning curve analysis in evaluating model reliability, beyond performance metrics alone. Future work should focus on optimizing model convergence and exploring hybrid approaches to improve clinical applicability in electrocardiogram signal classification.