Enhanced Hypertrophic Cardiomyopathy Detection Using Machine Learning on ECG Images

The work presented here is a comprehensive examination of machine learning algorithms for ECG image classification, focusing on their application to Hypertrophic Cardiomyopathy (HCM). The dataset used in this research consists of electrocardiogram (ECG) image data which has five categories, namely, Atrial Hypertrophy Block (AHB), COVID-19, Human-Machine Interface (HMI), MI with both infraction branch and opposite, and Normal. These categories are a benchmark for the categorization of ECG images that include conditions related to the heart's wellness such as HCM. For extracting features, a Convolutional Neural Network (CNN) was used to capture fine patterns in the ECG images. Moreover, when compared to other standard machine learning models, the CNN outperformed them in all performance metrics. The alternatives such as Support Vector Machine (SVM) with 52% accuracy, k-nearest Neighbors (k-NN) with 50%, and Logistic Regression with 55% are good benchmarks to compare against but not very useful than the Convolutional Neural Network (CNN). The results demonstrated a more effective use of CNN-based feature extraction in combination with classical machine learning for HCM diagnosis. Accurate identification and classification of HCM are crucial to timely management and improved patient outcomes.