Key points are not available for this paper at this time.
Abstract Background Coronary computed tomography angiography (CCTA) is known to be limited in assessing the hemodynamic significance of a coronary artery stenosis. A novel artificial intelligence (AI)-based CCTA-derived algorithm (AI-QCTischemia) aims to identify ischemic coronary artery disease (CAD) directly from CCTA images. Its incremental diagnostic value on top of clinical risk factors, coronary calcium score (CACS), and obstructive diameter stenosis (DS) for the prediction of myocardial ischemia remains unknown. Purpose To assess the incremental value of AI-QCTischemia for the prediction of ischemic CAD using a reference standard of hybrid CCTA/positron emission tomography (PET) perfusion imaging. Methods This was a cohort of 1448 symptomatic patients with suspected CAD having undergone hybrid CCTA/PET perfusion imaging (15O-H2O) to assess the presence of ischemic CAD. Four logistic regression models for the prediction of ischemia according to hybrid CCTA/PET perfusion imaging were assessed: 1) clinical: age, sex, hypertension, diabetes mellitus, smoking, typical angina pectoris; 2) clinical+CACS; 3) clinical+CACS+DS (i.e. visual DS ≥50%); 4) clinical+CACS+DS+AI-QCTischemia. Model performance was evaluated using area under the receiver-operating characteristic curves (AUC), and Delong’s method to compare AUCs between the models. Results 277/1448 (19.1%) patients had ischemic CAD according to hybrid CCTA/PET perfusion imaging. AUC (95% CI) for the clinical model was 0.75 (0.72-0.78), for the clinical+CACS model 0.86 (0.83-0.88), for the clinical+CACS+DS model 0.93 (0.91-0.95), and for the clinical+CACS+DS+AI-QCTischemia model 0.94 (0.92-0.95). Discriminatory ability improved significantly with each model from 1 to 4 with the model including AI-QCTischemia showing the overall highest performance (model 2 vs. model 1: p0.001; model 3 vs. model 2: p0.001; model 4 vs. model 3: p=0.011). Conclusion A novel AI-based CCTA-derived ischemia algorithm showed incremental diagnostic value on top of clinical risk factors, CACS, and obstructive stenosis on CCTA to predict to presence of ischemic CAD defined by the reference standard hybrid CCTA/PET perfusion imaging.AUCs for ischemic CAD
Bär et al. (Thu,) studied this question.