Impact of ECG Signal Preprocessing and Filtering on Arrhythmia Classification Using Machine Learning Techniques

Cardiac arrhythmias are heartbeat disorders in which the electrical impulses that coordinate the cardiac cycle malfunction. The heart's electrical activity is recorded using electrocardiography (ECG), a non-invasive method that helps diagnose several cardiovascular diseases. However, interpretation of ECG signals can be difficult due to the presence of noise, the irregularity of the heartbeat, and their nonstationary nature. Hence, the use of computational systems is required to support the diagnosis of cardiac arrhythmias. The main challenge in developing AI-assisted ECG systems is achieving accuracies suitable for application in clinical settings. Therefore, this paper introduces a software tool for classifying cardiac arrhythmias in ECG recordings that uses filtering, segmentation, and feature extraction of the QRS interval. We use the MIT-BIH Arrhythmia Database, which has 48 records of five different types of arrhythmias. We evaluate the data using supervised machine learning techniques such as k-Nearest Neighbors (KNN), Random Forest (RF), Multilayer Perceptron (MLP), and the Naive Bayesian classifier. This paper shows the impact of selecting and employing filtering and feature extraction methods on the performance of supervised machine learning algorithms compared with benchmark approaches.