Clinical Applicability of Machine Learning Models for Binary and Multi-Class Electrocardiogram Classification DOI Creative Commons

Daniel Nasef,

Demarcus Nasef,

Kennette James Basco

et al.

AI, Journal Year: 2025, Volume and Issue: 6(3), P. 59 - 59

Published: March 14, 2025

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

Language: Английский

Clinical Applicability of Machine Learning Models for Binary and Multi-Class Electrocardiogram Classification DOI Creative Commons

Daniel Nasef,

Demarcus Nasef,

Kennette James Basco

et al.

AI, Journal Year: 2025, Volume and Issue: 6(3), P. 59 - 59

Published: March 14, 2025

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

Language: Английский

Citations

1