Detection transformers reveal fiber details from complex brain scans
Fiber-Resolved Microstructure Quantification from Multi-Shell Diffusion MRI using Detection Transformers
Computer Vision and Pattern RecognitionMachine Learning
Summary
Looking inside the brain's white matter is tricky because it contains many tiny fibers packed together. Existing methods either find the direction of these fibers or measure their properties, but not both at once when the fiber structure is complex. The authors used a special AI model called a Detection Transformer to predict both the fiber directions and their details from standard brain scans. Their approach works well even when there are several different fiber groups mixed in the same spot.
What this means in practice
- •For medical imaging teams: Extract detailed fiber directions and microstructure from standard multi-shell diffusion MRI scans with flexible compartment number per voxel.
- •For neurotechnology developers: Incorporate fiber-resolved microstructure estimates into brain mapping tools to enhance diagnostic and research capabilities.
Authors
Sebastian Endt, Marcus Wirth, Johannes Reinhold Schlund, Marion Irene Menzel
Abstract
Fiber orientation and compartmental microstructure are central to the characterization of white matter tissue in diffusion MRI, yet existing methods either resolve fiber orientations without quantifying microstructure, or quantify microstructure while assuming a fixed number of compartments and a single fiber direction. Nonparametric approaches that recover both require tensor-valued diffusion encoding and computationally expensive Monte-Carlo inversion of an ill-posed inverse Laplace transform. We propose to reframe this problem as an object detection-like task, adopting the Detection Transformer (DETR) architecture to jointly predict mean diffusivity (MD), fractional anisotropy (FA), main fiber direction, and signal fraction for a variable number of compartments per voxel from standard multi-shell diffusion MRI with linear encoding. Hungarian matching during training resolves permutation invariance across compartments. We introduce mean Average Precision as a reproducible benchmark metric. Evaluated on synthetic test data with up to five compartments per voxel, our model achieves $R^2=0.95$ for MD, $R^2=0.88$ for FA, and a median angular error of 4.2°, with performance scaling naturally with compartmental signal fraction.