Toward optimal inline respiratory motion correction for in vivo cardiac diffusion tensor MRI using symmetric and inverse-consistent deformable image registration.

Abstract

Purpose: This study aims to develop a free-breathing cardiac DTI method with fast and robust motion correction.

Methods: Two proposed image registration-based motion correction (MOCO) strategies, MOCO_Naive and MOCO_Avg, were applied to diffusion-weighted images acquired with M2 diffusion gradients under free-breathing. The effectiveness of MOCO was assessed by tracking epicardium pixel positions across image frames. Resulting mean diffusivity (MD), fractional anisotropy (FA), and helix angle (HA) maps were compared against a previous low rank tensor based MOCO method (MOCO_LRT) in 20 healthy volunteers and two patients scanned at 3 T.

Results: Compared with the MOCO_LRT method, both proposed MOCO_Naive and MOCO_Avg methods generated slightly lower MD and helix angle transmurality (HAT) magnitude values, and significantly lower FA values. Moreover, both proposed MOCO methods achieved significantly smaller SDs of MD and FA values, and more smoothly varying helical structure in HA maps in healthy volunteers, indicating more effective MOCO. Elevated MD, decreased FA, and lower HAT magnitude were observed in two patients compared with healthy volunteers. Furthermore, the computing speed of image registration-based MOCO is twice as fast as the LRT method on the same dataset and same workstation.

Conclusion: This study demonstrates a fast and robust motion correction approach using image registration for in vivo free-breathing cardiac DTI. It improves the quality of quantitative diffusion maps and will facilitate clinical translation of cardiac DTI.