Deep learning based simultaneous measurement of flow-wall dynamics (DL-SFW)_experimental data

Several pathological phenomena are closely associated with vascular stiffness and interactions of blood flow and wall dynamics. However, conventional elastography and imaging techniques cannot easily measure local stiffness and analyze complicated interactions between multiple parameters. In this study, a new deep learning based simultaneous measurement of flow–wall dynamics (DL-SFW) is proposed by integrating and enhancing our two DL techniques for high-resolution velocimetry and strain measurement. The performance of DL-SFW is verified by comparing with conventional techniques for tissue-mimicking phantoms. The DL-SFW approach is found to improve relative errors in the measurements of velocity, wall shear stress (WSS), and strain with up to 4.6-fold, 15.1-fold, and 22.2-fold, respectively. After performance verification, in vivo feasibility is demonstrated by applying the DL-SFW to murine carotid artery under pathological conditions including aging and diabetes mellitus (DM). Mean flow velocities and mean WSS values of the aging and DM groups are lower than those of the control group. However, the strain values of the aging and DM groups are much smaller than that of the control group (p < 0.005). The mutual comparison of flow–wall dynamics and histological analysis results shows that the increase in vascular thickness and immunoreactive region is closely correlated with the regions with abnormal interactions between blood flow and wall dynamics.This study elucidates the excellent performance of DL-SFW in the simultaneous measurements of vascular stiffness and complicated flow–wall dynamics. Furthermore, these results provide useful information with high-resolution and accurate diagnosis of cardiovascular diseases.