Motivation: Eddy-current-induced distortions in diffusion MRI (dMRI) cause misalignment between volumes, disrupting downstream modeling and analysis. Current correction methods rely on traditional optimization, which is computationally intensive. Goal(s): To develop a deep-learning approach that efficiently corrects for eddy-current distortions in dMRI, offering a faster alternative to slow traditional methods. Approach: Two deep-learning models in sequence, both convolutional: 1) An image translator to restore correspondences between volumes. 2) A registration model to estimate the distortion parameters and apply correction. Results: The proposed method achieves distortion correction comparable to the widely-used FSL Eddy tool, but very rapidly at inference. Impact: Together with recently developed deep-learning susceptibility-induced distortion correction techniques, this work paves the way for real-time preprocessing of dMRI, facilitating its wider uptake in the clinic.
Legouhy et al. (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: