CARDIAC ABNORMALITIES DIAGNOSIS IN ELECTROCARDIOGRAMS USING CONVOLUTIONAL NEURAL NETWORK

Abstract

Cardiovascular diseases, primarily heart conditions, are a prominent global cause of death, making early
prediction vital. The cost-effective and noninvasive tool, Electrocardiogram (ECG), aids in detecting these diseases
by monitoring heart activity. To enhance prediction accuracy, deep learning techniques are employed to identify
four significant cardiac abnormalities: abnormal heartbeat, myocardial infarction, history of myocardial infarction,
and normal cases. The project combines transfer learning from deep neural networks like SqueezeNet and AlexNet
with a specialized CNN architecture. This approach is for extracting important features, improving predictions when
integrated with traditional machine learning algorithms. The proposed model stands out by delivering exceptional
performance, significantly advancing the prediction of medical conditions from images. It underscores the crucial
role of artificial intelligence in revolutionizing healthcare practices. The integrated Xception model elevates feature
extraction for cardiac abnormality detection from ECG images. Extracted features serve as inputs to machine
learning models, enhancing their ability to discern intricate patterns and anomalies. This amalgamation of advanced
feature extraction with robust machine learning algorithms contributes to the project's effectiveness in providing
accurate diagnoses. The streamlined user interactions through Flask with SQLite underscore the system's
practicality, offering secure signup, signin, and efficient testing for improved healthcare practices.