Detecting Sinus Bradycardia From ECG Signals Using Signal Processing And Machine Learning

Heart arrhythmias provide challenges for medical professionals due to their unique nature across individuals and the practice required to diagnose efficiently. With the aid of machine learning, however, this process can be improved to the point where diagnosis of a person's heartbeat can be done nearly instantaneously with no human interference. In this paper, I propose a Machine Learning (ML) model that is able to detect the occurrence or absence of sinus bradycardia in an Electrocardiogram (ECG) sample taken from a patient. While performing the standard signal processing pipeline, I evaluate the effectiveness of the features I extract, experimenting with Heart Rate Variability (HRV) metrics. I also test applying Principal Component Analysis (PCA) to the features in order to improve the performance of the ML models. Six models were created, both with and without PCA, with the aid of hyperparameter tuning to maximize accuracy and to ensure the efficient classification of selected data. The best performing model (K-Nearest Neighbors) was able to achieve an accuracy of 98%, precision of 95 %, and F1 Score of 96.7%. My results demonstrate the high potential, especially in clinical applications, in using electrocardiogram signals to diagnose heart arrhythmias.