Fiber-Resolved Microstructure Quantification from Multi-Shell Diffusion MRI using Detection Transformers
A Detection Transformer estimates multi-fiber diffusion MRI microstructure from standard multi-shell scans.
The paper treats fiber orientation and compartment microstructure in diffusion MRI as an object-detection task using a Detection Transformer. From standard multi-shell scans with linear encoding, the model jointly predicts mean diffusivity, fractional anisotropy, main fiber direction, and signal fraction for a variable number of compartments, with Hungarian matching during training. On synthetic data with up to five compartments per voxel, it reports R-squared of 0.95 for mean diffusivity, 0.88 for fractional anisotropy, and a median angular error of 4.2 degrees. Mean Average Precision is proposed as the benchmark metric.
- DETR jointly predicts diffusivity, anisotropy, fiber direction, and signal fraction.
- Hungarian matching handles a variable number of compartments per voxel.
- Synthetic tests report R-squared 0.95 for MD and 0.88 for FA.
- Median angular error is 4.2 degrees, scaling with signal fraction.
From the original article
1 min read at arxiv.orgFiber 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.
By Sebastian Endt · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.