Abstract
BACKGROUND AND OBJECTIVES: The hemodynamic evaluation of coronary stenoses undergoes a transition from wire-based invasive measurements to image-based computational assessments. However, fractional flow reserve (FFR) values derived from coronary CT angiography (CCTA) and angiography-based quantitative flow ratio have certain limitations in accuracy and efficiency, preventing their widespread use in routine practice. Hence, we aimed to investigate the diagnostic performance of FFR derived from the integration of CCTA and invasive angiography (FFR(CT-angio)) with artificial intelligence assistance in patients with stable coronary artery disease (CAD). METHODS: Forty stable CAD patients with 67 target vessels (50%-90% diameter stenosis) were included in this single-center retrospective study. All patients underwent CCTA followed by coronary angiography with FFR measurement within 30 days. Both CCTA and angiographic images were combined to generate a three-dimensional reconstruction of the coronary arteries using artificial intelligence. Subsequently, functional assessment was performed through a deep learning algorithm. FFR was used as the reference. RESULTS: FFR(CT-angio) values were significantly correlated with FFR values (r = 0.81, P < 0.001, Spearman analysis). Per-vessel diagnostic accuracy of FFR(CT-angio) was 92.54%. Sensitivity and specificity in identifying ischemic lesions were 100% and 88.10%, respectively. Positive predictive value and negative predictive value were 83.33% and 100%, respectively. Moreover, the diagnostic performance of FFR(CT-angio) was satisfactory in different target vessels and different segment lesions. CONCLUSIONS: FFR(CT-angio) exhibits excellent diagnostic performance of identifying ischemic lesions in patients with stable CAD. Combining CCTA and angiographic imaging, FFR(CT-angio) may represent an effective and practical alternative to invasive FFR in selected patients.