Machine Learning-based Biomarker Identification for Cardiovascular Disease Prediction: Combining X-ray with Electrocardiogram Analysis

Abstract

Cardiovascular disease (CVD) is a major health problem worldwide. As there currently is no cure for cardiovascular disease, early diagnosis is imperative in maintaining a high quality of life for patients. Traditional method of diagnosis for cardiovascular disease involves assessing a coronary artery calcium score (CACS) through computerized tomography (CT). However, this method cannot achieve early diagnosis of cardiovascular disease. This study aims to find a novel biomarker for cardiovascular disease using X-rays and electrocardiograms (ECG). A machine learning-based approach will be used to create a multimodal algorithm that can diagnose the severity of CVD through using CACS benchmarks. Numerous experimental results indicate that the proposed multimodal approach enhances the accuracy of diagnosing the severity of CVD. The proposed multimodal approach was evaluated with several state-of-the-art convolutional neural network architectures, resulting in an impressive accuracy rate of 93.8% on a large-scale CVD dataset comprised of 21,625 patient samples. These findings demonstrate the feasibility of utilizing biomarkers such as ECG and X-ray for diagnosing CVD.