Optimizing arrhythmia identification using chaos theory and deep learning analysis

Abstract Optimal heart function is crucial for quality life. Electrocardiograms are indispensable for diagnosing cardiac irregularities and analyzing heartbeat signals. This study unveils three novel machine learning (ML) techniques enhancing diagnostic precision. The first method scrutinizes raw ECG data and its time-series metrics, while the second incorporates historical patient health data for direct ECG classification. The third technique transforms ECG signals into insightful features, focusing on QRS complex waves. Uniquely combining these strategies, our research pioneers in advanced feature selection for cardiac diagnosis. Experiments were conducted to compare the integration of clinical data, QRS characteristics, and chaos preprocessing impact on the diagnosis. For classification assessment, five algorithms were utilized: Support Vector Machine, Decision Tree, Naïve Bayes, Neural Network, and Convolutional Neural Networks (CNNs). Implementing chaos theory, we converted QRS features into deterministic metrics. Notably, our Chaos-Enhanced CNN model exhibited outstanding performance, achieving a remarkable 99.8% accuracy rate, outshining other models.