DACO: Distortion/artefact correction for diffusion MRI data.

Area of Science:

• Medical Imaging
• Neuroimaging
• Biophysics

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) is susceptible to various artifacts.
• Susceptibility-induced distortions, eddy current-induced distortions, and head motion can compromise dMRI data quality.
• Current correction methods often require additional imaging sequences or complex processing.

Purpose of the Study:

• To introduce DACO, a novel registration-based algorithm for correcting common distortions and artifacts in dMRI.
• To demonstrate DACO's ability to improve dMRI data quality by leveraging routinely acquired anatomical images.
• To provide a method that minimizes the need for additional acquisitions in dMRI processing.

Main Methods:

• DACO synthesizes pseudo b0 and pseudo dMRI data from anatomical images (T1w/T2w) and diffusion models (DTI/MAP-MRI).
• It employs a registration-based approach, aligning real b0 to pseudo b0 to correct susceptibility artifacts and misalignment.
• Iterative registration of real dMRI data to pseudo dMRI data corrects eddy current distortions and head motion simultaneously.

Main Results:

• DACO accurately estimated model parameters on Human Connectome Project (HCP) data.
• Evaluations on clinical MRI scanner data showed effective reduction of artifacts.
• The method successfully corrected susceptibility-induced distortions, misalignment, eddy current distortions, and head motion.

Conclusions:

• DACO offers an effective, registration-based solution for correcting multiple dMRI artifacts.
• The algorithm's reliance on standard anatomical images makes it broadly applicable, especially in clinical settings.
• DACO is particularly beneficial for dMRI datasets lacking field maps or reverse phase-encoding images.