Effect of Tube Voltage on Diagnostic Performance of Fractional Flow Reserve Derived From Coronary CT Angiography With Machine Learning: Results From the MACHINE Registry
OBJECTIVE. Coronary CT angiography (CCTA)-based methods allow noninvasive estimation of fractional flow reserve (cFFR), recently through use of a machine learning (ML) algorithm (cFFRML). However, attenuation values vary according to the tube voltage used, and it has not been shown whether this significantly affects the diagnostic performance of cFFR and cFFRML. Therefore, the purpose of this study is to retrospectively evaluate the effect of tube voltage on the diagnostic performance of cFFRML. MATERIALS AND METHODS. A total of 525 coronary vessels in 351 patients identified in the MACHINE consortium registry were evaluated in terms of invasively measured FFR and cFFRML. CCTA examinations were performed with a tube voltage of 80, 100, or 120 kVp. For each tube voltage value, correlation (assessed by Spearman rank correlation coefficient), agreement (evaluated by intraclass correlation coefficient and Bland-Altman plot analysis), and diagnostic performance (based on ROC AUC value, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy) of the cFFRML in terms of detection of significant stenosis were calculated. RESULTS. For tube voltages of 80, 100, and 120 kVp, the Spearman correlation coefficient for cFFRML in relation to the invasively measured FFR value was ρ = 0.684, ρ = 0.622, and ρ = 0.669, respectively (p < 0.001 for all). The corresponding intraclass correlation coefficient was 0.78, 0.76, and 0.77, respectively (p < 0.001 for all). Sensitivity was 100.0%, 73.5%, and 85.0%, and specificity was 76.2%, 79.0%, and 72.8% for tube voltages of 80, 100, and 120 kVp, respectively. The ROC AUC value was 0.90, 0.82, and 0.80 for 80, 100, and 120 kVp, respectively (p < 0.001 for all). CONCLUSION. CCTA-derived cFFRML is a robust method, and its performance does not vary significantly between examinations performed using tube voltages of 100 kVp and 120 kVp. However, because of rapid advancements in CT and postprocessing technology, further research is needed.
|Keywords||coronary artery disease, coronary CT angiography, fractional flow reserve, machine learning, tube voltage|
|Persistent URL||dx.doi.org/10.2214/AJR.18.20774, hdl.handle.net/1765/118590|
|Journal||American Journal of Roentgenology|
De Geer, J. (Jakob), Coenen, A, Kim, Y.-H, Kruk, M. (Mariusz), Tesche, C. (Christian), Schoepf, U.J, … Persson, A. (Anders). (2019). Effect of Tube Voltage on Diagnostic Performance of Fractional Flow Reserve Derived From Coronary CT Angiography With Machine Learning: Results From the MACHINE Registry. American Journal of Roentgenology, 213(2), 325–331. doi:10.2214/AJR.18.20774