Outputs#

mpat.run writes to <output_dir>/mpat/, or to mpat-moments/ with moments=True, next to pyAFQ’s own outputs.

file

<sub>_bundle-<Name>_desc-density_map.nii.gz

expected steps of bundle streamlines in each voxel, summed over all seeds

..._desc-endpoints_map.nii.gz

expected bundle endpoints per grey-matter or edge voxel (two per streamline)

..._desc-visitodf_dwimap.nii.gz

visitation ODF: the directions the bundle’s streamlines travel in each voxel, as SH in the fODF’s basis (DIPY descoteaux07, lmax 8), in expected visits per steradian; it integrates to the density. Written where the density is at least 1e-3 of its peak

..._desc-seedprob_probmap.nii.gz

probability that the streamline seeded in each interface voxel belongs to the bundle

..._desc-start_mask, _desc-end_mask

the endpoint ROIs used (pyAFQ’s dist_to_atlas applied)

..._desc-include_mask, _desc-exclude_mask

the include ROIs used (labeled 1, 2, … in the definition’s order) and the exclude ROIs (without the other hemisphere), sealed (Bundle definitions: include, exclude and endpoint ROIs)

..._desc-core_path.tsv, _desc-core.trx

the bundle’s core: the cross-section ridge path, 100 nodes

..._desc-summary.json

settings, solve diagnostics, mass check

<sub>_desc-mpat_table.tsv

per bundle: expected streamlines, columns, steps, Krylov products, time

<sub>_desc-mpat_dseg.nii.gz

the tissue labels used: 1 WM, 2 GM, 3 CSF, 4 brain edge

<sub>_desc-chain_cache.npz

the transition matrix, reused when the inputs and settings match

Reading the summary#

Each bundle’s *_desc-summary.json holds:

  • expected_bundle_streamlines: the sum of the seed probabilities. Every seed contributes one streamline.

  • endpoint_mass_check: the total endpoint mass divided by twice the expected streamlines. It should be 1.

  • backward and forward: iterations, final residual, and, if the Krylov tail ran, where it started and how many products it took.

  • negative_mass_before_clip: nonzero only with moments, whose linear closure can produce small negative masses (see 12. Spatial moments).

  • solved_with: the other bundles that shared this bundle’s passes.

How the maps are defined#

The method behind each map is in How MPAT works, 13. What comes out:

  • the density is the reactive density summed over direction classes

  • the visitation ODFs keep the direction classes

  • the core is the ridge path of the density