A knowledge graph-based analytical model for mining clinical value of drug stress echocardiography for diagnosis, risk stratification and prognostic evaluation of coronary artery disease

Abstract

The three major techniques for clinically diagnosing coronary heart disease, including angina associated with myocardial ischemia, are coronary angiography, myocardial perfusion imaging, and drug stress echocardiography. Compared to the first two methods, which are invasive or involve the use of radionuclides, drug stress echocardiography is increasingly used in clinical practice due to its non-invasive, low-risk, and controllable nature, and wide applicability. We developed a novel methodology to demonstrate knowledge graph-based efficacy analysis of drug stress echocardiography as a complement to traditional meta-analysis. By measuring coronary flow reserve (CFR), we discovered that regional ventricular wall abnormalities (RVWA) and drug-loaded cardiac ultrasound can be used to detect coronary artery disease. Additionally, drug-loaded cardiac ultrasound can be used to identify areas of cardiac ischemia, stratify risks, and determine prognosis. Furthermore, adenosine stress echocardiography(ASE) can determine atypical symptoms of coronary heart disease with associated cardiac events through CFR and related quantitative indices for risk stratification. Using a knowledge graph-based approach, we investigated the positive and negative effects of three drugs - Dipyridamole, Dobutamine, and Adenosine - for coronary artery disease analysis. Our findings show that Adenosine has the highest positive effect and the lowest negative effect among the three drugs. Due to its minimal and controlled side effects, and high sensitivity for diagnosing coronary microcirculation disorders and multiple lesions, adenosine is frequently used in clinical practice.