An artificial intelligence-driven 3D-vectorcardiography technique for non-invasive prediction of obstructive coronary artery disease: a prospective study

Abstract

Cardiovascular disease, particularly coronary artery disease (CAD), is the leading cause of death in industrialized nations. Invasive coronary angiography is the diagnostic gold standard for ischemic heart disease but is costly, and complex. An innovative approach employs artificial intelligence (AI) in 3D-vectorcardiography for non-invasive identification of obstructive coronary lesions, integrating neural networks within a supervised learning framework. This study aimed to evaluate the accuracy of AI-driven 3D-vectorcardiography for noninvasive identification of obstructive CAD compared to invasive coronary angiography. In a prospective blinded study, 183 patients with possible CAD underwent AI-driven 3D-vectorcardiography before coronary angiography. A neural network AI algorithm calculated parameters, including a perfusion factor, to assess the probability of obstructive CAD. Investigators analyzing the AI-driven 3D-vectorcardiography and coronary angiographies were unaware of each other’s results. The clinical risk model had a receiver operating characteristic (ROC) area under the curve (AUC) of 0.617. Incorporating AI-driven 3D-vectorcardiography considerable improved prediction accuracy, achieving an AUC of 0.716. AI-driven 3D-vectorcardiography is a simple and effective diagnostic tool for enhancing the noninvasive detection of obstructive CAD. Further studies using fractional flow reserve (FFR) and microcirculation measurements are needed to better define its role in predicting cardiac ischemia and CAD.