Estimating Fiber Orientation Distribution Through Blockwise Adaptive Thresholding with Application to HCP Young Adults Data

• Activity Number: 26 - Statistics in Imaging: Student Award Session
• Type: Topic Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 11:50 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Seungyong Hwang* and Thomas Lee and Jie Peng and Debashis Paul
• Companies: University of California, Davis and University of California, Davis and University of California, Davis and University of California, Davis
• Keywords: Spherical deconvolution; James-Stein estimator; Neuroimaging; Diffusion Weighted Imaging; Nonparametric regression

Abstract

Due to recent technological advances, large brain imaging data sets can now be collected. Such data are highly complex so extraction of meaningful information from them remains challenging. Thus, there is an urgent need for statistical procedures that are computationally scalable and can provide accurate estimates that capture the neuronal structures and their functionalities. We propose a fast method for estimating the fiber orientation distribution (FOD) based on diffusion MRI data. This method models the observed dMRI signal at any voxel as a convolved and noisy version of the underlying FOD, and utilizes the spherical harmonics basis for representing the FOD, where the spherical harmonic coefficients are adaptively and nonlinearly shrunk by using a James-Stein-type estimator. To further improve the estimation accuracy by enhancing the localized peaks of the FOD, as a second step a super-resolution sharpening process is then applied. The resulting estimated FODs can be fed to a fiber tracking algorithm to reconstruct the white matter fiber tracts. We illustrate the overall methodology using both synthetic data and data from the Human Connectome Project.