In American journal of physiology. Heart and circulatory physiology
Coronary artery stenosis resistance (SR) is a key factor for non-invasive calculations of fractional flow reserve derived from coronary CT angiography (FFRCT). Existing computational fluid dynamics (CFD) methods, including three-dimensional (3D) computational and zero-dimensional (0D) analytical models, are usually limited by high calculation cost or low precision. In this study, we have developed a multi-input back propagation neural network (BPNN) that can rapidly and accurately predict coronary SR. A training dataset including 3,028 idealized anatomic coronary artery stenosis models was constructed for 3D CFD calculation of SR with specific blood flow boundaries. Based on 3D calculation results, we established a BPNN whose input is geometric parameters and blood flow, while output is SR. Then a test set (324 cases) was constructed to evaluate the performance of the BPNN model. To verify the validity and practicability of the network, BPNN prediction results were compared with 3D CFD and 0D analytical model results from patient-specific models. For test set, the mean square error (MSE) between CFD and prediction results was 2.97%, linear regression analysis indicating a good correlation between the two (p<0.001). For 30 patient-specific models, the MSE of BPNN and the 0D model were 3.26% and 9.7%, respectively. The calculation time for BPNN and the 3D CFD model for 30 cases was about 2.15 seconds and 2 hours, respectively. The present results demonstrate the practicability of using deep learning methods for fast and accurate predictions of coronary artery SR. Our study represents an advance in non-invasive calculations of FFRCT.
Sun Hao, Liu Jincheng, Feng Yili, Xi Xiaolu, Xu Ke, Zhang Liyuan, Liu Jian, Li Bao, Liu Youjun
Coronary artery, Deep learning, Fractional Flow Reserve, Stenosis resistance