Diagnostic Performance of Machine Learning Based CT-FFR in Detecting Ischemia in Myocardial Bridging and Concomitant Proximal Atherosclerotic Disease

Abstract

BackgroundThe diagnostic performance of coronary computed tomography angiography-derived fractional flow reserve (CT-FFR) in detecting ischemia in myocardial bridging (MB) has not been investigated to date. MethodsThis retrospective multicentre study included 104 patients with left anterior descending MBs. MB was classified as either superficial or deep, short, or long, whereas all MB vessels were further divided into <50%, 50% to 69%, and ≥70% groups, according to proximal lumen stenosis on invasive coronary angiography. Diagnostic performance and receiver operating characteristics (ROC) of CT-FFR to detect lesion-specific ischemia was assessed on a per-vessel level, using invasive FFR as reference standard. Intraclass correlation coefficient (ICC) and Bland-Altman plots were used for agreement measurement. ResultsForty-eight MB vessels (46.2%) showed ischemia by invasive FFR (≤0.80). Sensitivity, specificity, and accuracy of CT-FFR to detect functional ischemia were 0.96 (0.85 to 0.99), 0.84 (0.71 to 0.92), and 0.89 (0.81 to 0.94), respectively, in all MB vessels. There were no differences in diagnostic performance between superficial and deep MB or between short and long MB (all P > 0.05). The accuracy of CT-FFR was 0.96 (0.85 to 0.99) in ≥70% stenosis, 0.82 (0.67 to 0.91) in 50% to 69% stenosis, and 0.89 (0.51 to 0.99) in <50% stenosis (P = 0.081). Bland-Altman analysis showed a slight mean difference between CT-FFR and invasive FFR of 0.014 (95% limit of agreement, –0.117 to 0.145). The ICC was 0.775 (95% confidence interval, 0.685-0.842, P < 0.001). ConclusionsCT-FFR demonstrated high diagnostic performance for identifying functional ischemia in vessels with MB and concomitant proximal atherosclerotic disease when compared with invasive FFR. However, the clinical use of CT-FFR in patients with MB needs further study for stronger and more robust results.