Diffusion MRI probes brain tissue architecture, but crossing white matter fibers confound standard analysis. Current fixel-based methods rely on peaks of fiber orientation distributions from constrained spherical deconvolution, which are biased by nearby fiber spreads and fail at angular separations below roughly 40–50 degrees.

Researchers propose FISSILE, a constrained nonlinear optimization that treats each voxel as a mixture of single-fiber distributions. In simulations, FISSILE recovered fibers at 40° separation with a volume fraction of 0.1 at an SNR of 30, while MRtrix3's fod2fixel was limited to 60° separation under the same conditions.

On real data from three Human Connectome Project subjects and an ex vivo raccoon brain, FISSILE identified at least one crossing in 85.8% of human white matter voxels and 79.4% of raccoon white matter voxels, compared with 30.1% and 44.5% for fod2fixel. The number of crossings was mildly sensitive to noise when the same human subject was scanned twice at different signal-to-noise ratios.

The authors conclude that modeling the crossing fiber phenomenon as an unmixing problem increases sensitivity to individual fibers for more precise characterization of anatomy. The work is available as a preprint on arXiv.