.png){kind=icon}
.png){kind=icon}
Return
A Wasserstein-Space-Based Framework for Processing Fiber Orientation Geometry in Diffusion MRI
.png){kind=icon}
PRE
AI
DOI：10.1109/TMI.2025.3595367.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The fiber orientation distribution (FOD) function is an advanced model for high angular resolution diffusion MRI, capable of representing complex crossing or fanning fiber geometries. However, the intricate mathematical structures of FOD functions pose significant challenges for data processing and analysis. Current frameworks often fail to consider fiber bundle rotation information among FOD peaks, leading to improper data processing, such as inaccurate FOD interpolation and, consequently, anatomically incorrect fiber tracking. This paper presents a novel Wasserstein-space-based framework for processing and analyzing FOD functions that systematically considers fiber-bundle-specific geometry. Our approach begins with a spherical deconvolution method to accurately detect and decompose FOD functions into single-peak lobes. These single-peak lobes are then embedded into the Wasserstein-space, where a new metric for FOD functions is defined, capable of handling rotations among peak lobes. We introduce a geometry-aware clustering method to regroup the single-peak lobes for further bundle-specific FOD processing. The proposed framework is applied to the essential task of FOD interpolation, computed as the Barycenter of the new metric, with a fast approximation method for efficient computation. Experiments conducted on synthetic data, as well as datasets from the Human Connectome Project (HCP) and the Alzheimer’s Disease Neuroimaging Initiative (ADNI), demonstrate that our framework effectively handles complex fiber geometries, provides anatomically meaningful FOD interpolations, and significantly enhances the performance of FOD-based tractography.
Keywords:
Diffusion MRI
fiber orientation distribution
Wasserstein-space
tractography
interpolation
Journal
IF:
9.8
Papers:
6.2K
Citations:
3.7W