Early Diagnosis of Heart Conditions Using AI-Driven on Electronic Health Data

Abstract

Heart failure, atherosclerosis, coronary artery disease, cardiomyopathies, arrhythmic disorders, valve diseases, and other cardiovascular diseases (CVDs) are the main causes of illness and death around the world. This work presents a practical method that uses deep learning to detect cardiac problems early using electrocardiogram (ECG) readings. It uses the ECG Heartbeat Categorization Dataset of Kaggle based on the MIT-BIH Arrhythmia Database which has more than 109,000 labeled ECG segments in five classes of heartbeats. Extensive data preprocessing is done, such as data cleaning, Z score normalization, one-hot encoding, and recursive feature elimination (RFE) of feature selection. As a measure to deal with a sharp disparity in the classes, the ADASYN oversampling method is used, with a balanced dataset being obtained. After that, a Recurrent Neural Network (RNN) is trained on the time-dependent nature of electrocardiogram (ECG) signals in order to identify heartbeats accurately. The accuracy (ACC), precision (PRE), recall (REC), F1-score (F1), and loss are the measures used to assess the performance of the model. With scores of 96.06% for accuracy, 90.0% for precision, 93.1% for recall, and 91.0% for F1-score, the suggested RNN model shows good performance, suggesting balanced and reliable categorization. Early cardiac problem detection and clinical decision-making may benefit from the proposed RNN technique, since it beats both conventional ML algorithms and rival deep learning architectures.