Field-of-view extension for brain diffusion MRI via deep generative models.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Diffusion Magnetic Resonance Imaging (dMRI)

Background:

• Incomplete field of view (FOV) in dMRI scans significantly hinders volumetric and bundle analyses of brain microstructure and connectivity.
• Discarding valuable dMRI data with incomplete FOV limits subsequent tractography analyses.
• There is a need for methods to impute missing data in dMRI scans to preserve data integrity and enable comprehensive analysis.

Purpose of the Study:

• To develop a novel method for imputing missing slices in dMRI scans with incomplete FOV.
• To enhance whole-brain tractography accuracy using dMRI data with imputed complete FOV.
• To provide an alternative to discarding valuable dMRI data affected by incomplete FOV.

Main Methods:

• A deep generative model framework was proposed to estimate absent brain regions in dMRI scans.
• The model learns diffusion characteristics from diffusion-weighted images (DWIs) and anatomical features from structural images.
• This enables efficient imputation of missing DWI slices within the incomplete FOV.

Main Results:

• The proposed framework demonstrated sufficient imputation performance on the Wisconsin Registry for Alzheimer's Prevention (WRAP) and National Alzheimer's Coordinating Center (NACC) datasets.
• Tractography accuracy was significantly improved, evidenced by an increased average Dice score for 72 tracts on both datasets.
• The method successfully repaired corrupted dMRI data with incomplete FOV.

Conclusions:

• The developed framework effectively imputes missing data in brain dMRI scans with incomplete FOV.
• The imputation improves whole-brain tractography, offering a solution for repairing corrupted neuroimaging data.
• This approach enhances the accuracy of analyzing brain bundles, particularly relevant for conditions like Alzheimer's disease.