Multimodal validation of patient-specific intraventricular flow simulations from 4D echocardiography

The combination of refined medical imaging techniques and computational fluid dynamics (CFD) models has enabled the study of complex flow behavior on a highly regional level. Recently, we have developed a platform for patient-specific CFD modelling of blood flow in the left ventricle (LV), with input data and required boundary conditions acquired from 4D echocardiography. The platform robustness has been evaluated with respect to input variable variations, but for any clinical implementation model flow validation is essential. Therefore, the aim of this study is to evaluate the accuracy of the patient-specific CFD model against multimodal image-based flow measurements. For the validation, 4D echocardiography was acquired from two healthy subjects, from which LV velocity fields were simulated. In-vivo flows from the same two subjects were then acquired by pulsed wave (PW) Doppler imaging over both LV-valves, and by cine phase-contract magnetic resonance imaging (PC-MRI) at eight defined anatomical planes in the LV. By fusing PC-MRI and the ultrasound acquisitions using a three-chamber alignment algorithm, simulated and measured flows were quantitatively compared. General flow pattern correspondence was observed, with a mean error of 1.4 cm/s and root mean square deviation of 5.7 cm/s for all measured PC-MRI LV-planes. For the PW-Doppler comparison, a mean error of 3.6 cm/s was reported. Overall, the following work represents a validation of the proposed patient-specific CFD platform, and the agreement with clinical data highlight the potential for future clinical use of the models.