CT morphological index provides incremental value to machine learning based CT-FFR for predicting hemodynamically significant coronary stenosis

Aims: To study the diagnostic performance of the ratio of Duke jeopardy score (DJS) to the minimal lumen diameter (MLD) at coronary computed tomographic angiography (CCTA) and machine learning based CT-FFR for differentiating functionally significant from insignificant lesions, with reference to fractional flow reserve (FFR). Methods and results: Patients who underwent both coronary CTA and FFR measurement at invasive coronary angiography (ICA) within 2 weeks were retrospectively included in our study. CT-FFR, DJS/MLDCT ratio, along with other parameters, including minimal luminal area (MLA), MLD, lesion length (LL), diameter stenosis, area stenosis, plaque burden, and remodeling index of lesions, were recorded. Lesions with FFR <= 0.8 were considered to be functionally significant. One hundred and twenty-nine patients with 166 lesions were ultimately included for analysis. The LL, diameter stenosis, area stenosis, plaque burden, DJS and DJS/MLDCT ratio were all significantly longer or larger in the group of FFR <= 0.8 (p < 0.001 for all), while smaller MLA, MLD and CT-FFR value were also noted (p < 0.001 for all). CT-FFR and DJS/MLDCT ratio showed the largest AUC among all single parameters (AUC = 0.85 and AUC = 0.83, respectively; p < 0.001 for both) for diagnosing functionally significant stenosis. Combining CT-FFR and DJS/MLDCT ratio provided incremental value for discrimination between flow-limiting and non-flow-limiting coronary lesions and yielded the best diagnostic performance (accuracy of 83.7%). Conclusions: The combination of ML-based CT-FFR and DJS/MLDCT allows accurate non-invasive discrimination between flow-limiting and non-flow-limiting coronary lesions. (c) 2018 Elsevier B.V. All rights reserved.