Deep learning-based prediction of coronary artery stenosis resistance.

Coronary artery stenosis resistance (SR) is a key factor for noninvasive calculations of fractional flow reserve derived from coronary CT angiography (FFRCT). Existing computational fluid dynamics (CFD) methods, including three-dimensional (3-D) computational and zero-dimensional (0-D) 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 data set including 3,028 idealized anatomic coronary artery stenosis models was constructed for 3-D CFD calculation of SR with specific blood flow boundaries. Based on 3-D calculation results, we established a BPNN whose input is geometric parameters and blood flow, whereas 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 3-D CFD and 0-D 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 0-D model were 3.26 and 9.7%, respectively. The calculation time for BPNN and the 3-D CFD model for 30 cases was about 2.15 s and 2 h, 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 noninvasive calculations of FFRCT.NEW & NOTEWORTHY This study developed a multi-input back-propagation neural network (BPNN) that can be used to predict coronary artery stenosis resistance by inputting vascular geometric parameters and blood flow. Compared with previous studies, the network developed in this study can accurately and rapidly predict coronary artery stenosis resistance, which can not only meet clinical requirements but also reduce the cost of calculation duration. This study contributes to the noninvasive methods for the numerical calculation of fractional flow reserve derived from coronary CT angiography (FFRCT) and indicates that this technique can potentially be used for evaluating myocardial ischemia.

Accurate Calculation of FFR Based on a Physics-Driven Fluid-Structure Interaction Model.

Background: The conventional FFRct numerical calculation method uses a model with a multi-scale geometry based upon CFD, and rigid walls. Therefore, important interactions between the elastic vessel wall and blood flow are not routinely considered. Changes in the resistance of coronary microcirculation during hyperaemia are likewise not typically incorporated using a fluid-structure interaction (FSI) algorithm. It is likely that both have resulted in FFRct calculation errors. Objective: In this study we incorporated both the influence of vascular elasticity and coronary microcirculatory structure on FFR, to improve the accuracy of FFRct calculation. Thus, in this study, a physics-driven 3D-0D coupled model including fluid-structure interaction was established to calculate accurate FFRct values. Methods: Based upon a novel geometric multi-scale modeling technology, a FSI simulation approach was used. A lumped parameter model (0D) was used as the outlet boundary condition for the 3D FSI coronary artery model to incorporate physiological microcirculation, with bidirectional coupling between the two models. Results: The accuracy, sensitivity, specificity, and both positive and negative predictive values of FFRDC calculated based upon the coupled 3D-0D model were 86.7, 66.7, 84.6, 66.7, and 91.7%, respectively. Compared to the calculated value using the basic CFD model (MSE = 5.9%, accuracy rate = 80%), the FFRCFD calculated based on the coupled 3D-0D model has a smaller MSE of 1.9%. Conclusion: The physics-driven coupled 3D-0D model that incorporates fluid-structure interactions not only consider the influence of the elastic vessel wall on blood flow, but also provides reliable microvascular resistance boundary conditions for the 3D FSI model. This allows for a calculation that is based upon conditions that are closer to the physiological environment, and thus improves the accuracy of FFRct calculation. It is likely that more accurate information will provide an enhanced recommendation regarding percutaneous coronary intervention (PCI) in the clinic.

Effect of the Coronary Arterial Diameter Derived From Coronary Computed Tomography Angiography on Fractional Flow Reserve.

BACKGROUND: Fractional flow reserve (FFR) is considered to be the criterion standard for the clinical diagnosis of functional myocardial ischemia. In this study, we explored the effect of the coronary arterial diameter derived from coronary computed tomography angiography on FFR. METHOD: We retrospectively reviewed the clinical information of 131 patients with moderate coronary artery stenosis. To compare the mean diameter of stenotic vessels, patients were divided into ischemic and nonischemic groups. According to the clinical statistics of the diameter of the ischemic group and the nonischemic group, we established 8 ideal models of coronary artery diameter of 4 mm (40%, 50%, 60%, and 70% stenosis) and diameter of 3 mm (40%, 50%, 60%, and 70% stenosis). Two sets of numerical simulation experiments were carried out: experiment 1 evaluated the variation rate of CT-based computation of non-invasive fractional flow reserve (FFRCT) with vessel diameters of 4 mm and 3 mm under different stenosis rates, and experiment 2 explored the variation of FFRCT with vessel diameters of 4 mm and 3 mm under different cardiac outputs. We simulated changes in the flow of narrow blood vessels by changes in cardiac output. RESULTS: According to clinical statistics, the mean ± SD diameter of stenotic vessels in the ischemic and nonischemic groups was 3.67 ± 0.77 mm and 3.31 ± 0.64 mm (P < 0.05 for difference), respectively. In experiment 1, the FFRCT of coronary with a diameter of 4 mm was 0.86, 0.80, 0.66, and 0.35, and that with a diameter of 3 mm was 0.90, 0.84, 0.71, and 0.50, respectively. In experiment 2, the FFRCT of the coronary vessel diameter of 4 mm was 0.84, 0.80, 0.76, and 0.72, respectively. The FFRCT coronary vessels with a diameter of 3 mm were 0.87, 0.84, 0.80, and 0.76, respectively. CONCLUSIONS: As the stenosis increases, compared with narrow blood vessel of small diameter, the narrow blood vessel with larger diameter is accompanied by faster flow rate changes and is more prone to ischemia.

Model-based evaluation of local hemodynamic effects of enhanced external counterpulsation.

BACKGROUND AND OBJECTIVES: The treatment benefits of enhanced external counterpulsation (EECP) heavily depends on hemodynamics. Global hemodynamics of EECP can cause blood flow redistribution in the circulatory system whereas local hemodynamic effects act on vascular endothelial cells (VECs). Local hemodynamic effects of EECP on VECs are important in the treatment of atherosclerosis, but currently cannot be not evaluated. Herein we aim to establish evaluation models of local hemodynamic effects based on the global hemodynamic indicators. METHODS: We established 0D/3D geometric multi-scale hemodynamic models of the coronary and cerebral artery of two healthy individuals to calculate the global hemodynamic indicators and the local hemodynamic effects. Clinical EECP trials were performed to verify the accuracy of the multi-scale hemodynamic model. The global hemodynamic indicators included diastolic blood pressure/systolic blood pressure (Q = D/S), mean arterial pressure (MAP), internal carotid artery flow (ICAF) and cerebral blood flow (CBF), whereas local hemodynamic effects focused on time-averaged wall shear stress (TAWSS). The correlation between these indicators was analyzed via Pearson correlation coefficient. Significantly related indicators were selected for curve-fitting to establish evaluation models of the coronary and cerebral artery. Moreover, clinical data of a coronary heart disease patient and a cerebral ischemic stroke patient were collected to verify the effectiveness of the application of the established evaluation models to real patients. RESULTS: For coronary artery, TAWSS was correlated to Q = D/S and ICAF (P < 0.05), whereas for cerebral artery, TAWSS was correlated to MAP and CBF (P < 0.05). The mean square error (MSE) between the evaluated values using evaluation model and the calculated values using 0D/3D model of TAWSS of the coronary and cerebral artery were 5.4% and 1.0%, respectively. The MSE of evaluation model applied to real patients was greater than that applied to healthy individuals, but within an acceptable range. CONCLUSIONS: The presented error demonstrated validity and accuracy of the evaluation models in clinical patients. Based on the evaluation models, global hemodynamic indicators could be used to evaluate the local hemodynamic effects under the current counterpulsation mode. With TAWSS range of 4-7 Pa as the target range, EECP strategies can further be optimized.