Study on the Diagnostic Accuracy of FFR Deep Learning Method in Coronary Artery CT Angiography Calculation
Objective This study aims to evaluate the diagnostic accuracy of deep learning methods for calculating fractional flow reserve(FFR)in coronary artery CT angiography,in order to provide reference for clinical diagnosis and treatment.Methods This is a single center Prospective cohort study,63 patients participated in the diagnostic performance evaluation of deep learning FFR.In order to evaluate the ischemic risk of coronary artery stenosis,an automatic quantification method for the three-dimensional geometric shape of coronary arteries and a deep learning based FFR prediction method were proposed.Evaluate the diagnostic performance of deep learning FFR using linear FFR as a reference standard.Determine the main evaluation factors using the area under the subject operating characteristic curve(AUC)analysis.Results For each patient level,deep learning FFR showed higher diagnostic performance in determining ischemic related lesions with an area under the curve of 0.928 compared to CTA stenosis severity of 0.664 when the critical value measured by reference FFR was ≤ 0.8.FFR of deep learning was correlated with FFR(R=0.686,P＜0.001),and the average difference was-0.006±0.0091(P=0.619).The accuracy,sensitivity,specificity,positive predictive value,and negative predictive value of the secondary evaluation were 87.3％,97.14％,and 95.45％,respectively.Conclusion Deep learning FFR is a new method that can effectively evaluate the functional significance of coronary artery stenosis.
NETL
NSTL
万方数据
维普
