Artificial Intelligence Techniques for Cardiovascular Disease Classification using 12-Lead ECG

Abstract

Cardiovascular disorders, including atrial fibrillation and bundle branch blockages, have a significant impact on global health and are associated with higher mortality rates. The accurate categorization of these disorders through ECG data is essential in order to improve the results for patients. This investigation addresses the challenge of classifying major cardiac conditions by utilizing seven ECG datasets and employing machine learning techniques. We seek to classify patients exhibiting atypical ECGs through advanced methodologies. We combined datasets, incorporating 12 leads, and utilized preprocessing techniques to enhance data quality. By employing various models, such as CNN, SVM, KNN, random forest, and logistic regression, our CNN combined with logistic regression attained an accuracy of $94 \%$ and a sensitivity of $94 \%$ on a test set comprising 1500 ECGs. This comprehensive approach minimizes false negatives and enhances diagnostic precision. The integration of multiple datasets and sophisticated preprocessing improves the dependability of our findings, highlighting the importance of data quality and comprehensive analysis in cardiac diagnostics. The results demonstrate how machine learning can improve cardiac diagnoses through thorough data integration and analysis.