Coronary CT Angiography–derived Fractional Flow Reserve: Machine Learning Algorithm versus Computational Fluid Dynamics Modeling

Abstract

Purpose To compare two technical approaches for determination of coronary computed tomography (CT) angiography-derived fractional flow reserve (FFR)-FFR derived from coronary CT angiography based on computational fluid dynamics (hereafter, FFR) and FFR derived from coronary CT angiography based on machine learning algorithm (hereafter, FFR)-against coronary CT angiography and quantitative coronary angiography (QCA). Materials and Methods A total of 85 patients (mean age, 62 years ± 11 [standard deviation]; 62% men) who had undergone coronary CT angiography followed by invasive FFR were included in this single-center retrospective study. FFR values were derived on-site from coronary CT angiography data sets by using both FFR and FFR. The performance of both techniques for detecting lesion-specific ischemia was compared against visual stenosis grading at coronary CT angiography, QCA, and invasive FFR as the reference standard. Results On a per-lesion and per-patient level, FFR showed a sensitivity of 79% and 90% and a specificity of 94% and 95%, respectively, for detecting lesion-specific ischemia. Meanwhile, FFR resulted in a sensitivity of 79% and 89% and a specificity of 93% and 93%, respectively, on a per-lesion and per-patient basis (P = .86 and P = .92). On a per-lesion level, the area under the receiver operating characteristics curve (AUC) of 0.89 for FFR and 0.89 for FFR showed significantly higher discriminatory power for detecting lesion-specific ischemia compared with that of coronary CT angiography (AUC, 0.61) and QCA (AUC, 0.69) (all P < .0001). Also, on a per-patient level, FFR (AUC, 0.91) and FFR (AUC, 0.91) performed significantly better than did coronary CT angiography (AUC, 0.65) and QCA (AUC, 0.68) (all P < .0001). Processing time for FFR was significantly shorter compared with that of FFR (40.5 minutes ± 6.3 vs 43.4 minutes ± 7.1; P = .042). Conclusion The FFR algorithm performs equally in detecting lesion-specific ischemia when compared with the FFR approach. Both methods outperform accuracy of coronary CT angiography and QCA in the detection of flow-limiting stenosis.

Impact of machine-learning CT-derived fractional flow reserve for the diagnosis and management of coronary artery disease in the randomized CRESCENT trials

“…CT-derived fractional flow reserve (CT-FFR) can compute FFR values from standard CCTA images without requiring additional testing and radiation exposure, and has shown a good correlation and agreement with invasive FFR in several studies [4][5][6][7]. On-site CT-FFR software enables CT-FFR analyses on standard workstation without transferring CT images [8,9]. Additionally, onsite CT-FFR computation through interpretation of anatomical features (based on machine-learning algorithms) can, in contrary to computational fluid dynamic-based algorithms, be performed in several seconds [8,10].…”

Putting machine learning into motion: applications in cardiovascular imaging

“…24 The technique also depends on the assumptions of a pre-specified model of flow physiology, but recently a preliminary study has shown that deep learning algorithms perform equally well as CFD in the detection and quantification of flow-limiting disease. 25…”

Comparison of Machine Learning Computed Tomography-Based Fractional Flow Reserve and Coronary CT Angiography-Derived Plaque Characteristics with Invasive Resting Full-Cycle Ratio

“…The on-site prototype software calculated blood flow and showed the hyperemic state in the coronary vessels based on patient-specific physiological conditions and flow dynamic models. The technical specification of the used CT-FFR ML algorithm is described in detail in a previously published study [8]. After on-site calculation, the CT-FFR ML software created a patient-specific anatomic color-coded 3-dimensional mesh of the coronary artery tree and aortic root.…”

“…Several studies have shown a significant correlation between non-invasive CT-FFR and invasive fractional flow reserve (FFR) [2,5,6]. Using an on-site prototype with a machine-learning algorithm (CT-FFR ML ) to reduce computation time was the consequent next step in technical development [7][8][9][10]. CT-FFR ML was compared in previously published studies to the invasive gold standards invasive FFR [8] and instantaneous wave-free ratio (iwFR) [11] and showed promising results.…”