Incremental value of a CCTA-derived AI-based ischemia algorithm over standard CCTA interpretation for coronary ischemia in patients with suspected coronary artery disease

Abstract

Abstract Background Coronary computed tomography angiography (CCTA) is a widely used non-invasive test to diagnose coronary artery stenosis in patients with suspected coronary artery disease (CAD). Positive predictive value for ischemia has been suboptimal using clinical reading. A novel AI-based ischemia algorithm applying machine learning that considers an array of atherosclerotic and vascular morphology features from CCTA images has been suggested to improve the diagnostic performance for ischemia. Purpose This study aims to evaluate the incremental diagnostic performance of CCTA derived AI-based ischemia algorithm in evaluating myocardial ischemia. Methods Patients with suspected CAD referred for CCTA at Turku University Hospital from February 2007 to December 2016 were analysed. The AI-based ischemia algorithm was calculated by analysts blinded for patient characteristics and clinical outcomes using typically-acquired CCTA images. In diagnostic reading significant stenosis defined by a stenosis > 50% on CCTA. The reference standard was ischemia detected by stress 15O-H2O positron emission tomography (PET). Results A total of 1445 patients (mean age 63 ± 9 years, 46% male) were included in this analysis. Myocardial ischemia by PET was detected in 384 (27%) patients while the other patients had either normal PET or had non-obstructive CAD by CCTA. A positive AI-based ischemia algorithm was found in 431 (30%) patients. The diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the AI-based ischemia algorithm for the assessment of myocardial ischemia were 83%, 74%, 86%, 66%, and 90%, respectively. Compared with diagnostic reading of CCTA (area under the receiver operating characteristic curve [AUC], 0.853; 95%CI 0.832-0.874), the model including clinical reading and AI-based ischemia algorithm demonstrated superior discrimination of myocardial ischemia (AUC 0.883; 95%CI 0.863-0.904, p-value <0.001). (Figure) Conclusions A novel AI-based ischemia algorithm derived from CCTA improves diagnostic accuracy compared with clinical reading of CCTA in assessing myocardial ischemia by PET perfusion imaging in patients with suspected CAD.Figure