Key points are not available for this paper at this time.
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); however, it does not provide information of the functional ischemia. A novel artificial intelligence -guided quantitative CT ischemia algorithm (AI-QCTischemia) has been established, which comprises a machine-learned method using features of atherosclerosis and vascular morphology from CCTA images to identify the likelihood of myocardial ischemia. Purpose The aim of this study is to investigate the diagnostic performance of AI-QCTischemia over standard CCTA interpretation in detecting myocardial ischemia using positron emission tomography (PET) perfusion imaging as a reference standard. Methods Patients with suspected CAD referred for CCTA at our University Hospital from February 2007 to December 2016 were analysed. The AI-QCTischemia algorithm was applied by analysts blinded for patient characteristics and clinical outcomes using the CCTA images to detect probability of myocardial ischemia per patient level. In diagnostic reading significant stenosis was defined as 50% on CCTA. The reference standard was ischemia detected by stress 15O-H2O PET. Results A total of 1746 patients (mean age 62 ± 10 years, 44% male) were included. Myocardial ischemia by PET was detected in 325 (19%) patients while the other patients had either normal PET or had non-obstructive CAD by CCTA. AI-QCTischemia was found positive in 430 (25%) patients. The diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the AI-QCTischemia algorithm for the assessment of myocardial ischemia were 87%, 81%, 88%, 61%, and 95%, respectively. Compared with the clinical visual reading of CCTA, AI-QCTischemia demonstrated higher specificity and positive predictive value but somewhat lower sensitivity and negative predictive value. The accuracy of AI-QCTischemia was 87% while it was 86% with visual reading of CCTA. (Figure 1) Compared with diagnostic reading of CCTA (area under the receiver operating characteristic curve AUC, 0.868), the model including clinical reading and AI-QCTischemia algorithm demonstrated superior discrimination of myocardial ischemia (AUC 0.899, p-value 0.001). (Figure 2) Conclusions A novel AI-based ischemia algorithm derived from CCTA has a high diagnostic accuracy for predicting myocardial ischemia defined by PET perfusion imaging in symptomatic patients with suspected CAD. Moreover, combination of novel AI-based ischemia algorithm and clinical reading of CCTA demonstrated that incremental diagnostic accuracy for predicting diagnostic accuracy.Diagnostic accuracy of CCTA and AI-QCT AUC for myocardial ischemia on PET
Nabeta et al. (Thu,) studied this question.