Skip to content

MPAT live bundle

The template from the MPAT guide: solve one bundle with MPAT and watch its density build up as the solver runs. Download the workflow.

MPAT live bundle diagram

What this workflow does

It solves one bundle with MPAT from the files a mpat --save-transitions odx run writes, and shows the bundle's density building up in 3-D while the solver converges. The MPAT guide linked above explains how to make those files and how to fill in the template's {stem}, {bundle} and {t1w} placeholders.

A few things to look at in the diagram:

  • The four ROI masks become one tracking plan.
  • Each mask is read as a NIfTI and thresholded into a voxel mask.
  • Add End Region is chained twice: the first region added is the bundle's start, the second its end.
  • Add ROI adds the include mask and Add ROA the exclude mask.
  • The chain and the plan meet at MPAT Tractography.
  • The transitions ODX feeds MPAT Transition Probabilities, which loads the chain MPAT built.
  • The finished plan comes in on the tractography node's Tracking Plan port.
  • The glow is the Visitation Density output.
  • A Volume display draws it with direct volume rendering and the Hot colormap.
  • While the solver runs, that output shows the latest density frame.
  • Display nodes start quiet. The flux streamlines and ODF glyphs are hidden, and the ROIs are drawn at 12% opacity, so the glow stands out. Turn them up in each node's inspector.

Methods boilerplate

Rendered from the runtime methods-report generator for this workflow. Citation keys resolve to footnotes via mkdocs-bibtex; running the raw markdown through Pandoc + the project's citations.bib gives the same result for an exported paper.

Methods

Analytic voxel-to-voxel transition probabilities and the 26-direction-class absorbing Markov chain were constructed with MITTENS and MPAT 1 using a 35° turning cone, 1.00-voxel step size, double-ODF look-ahead, FOD threshold 0.05, FOD exponent 1.00, stopping mass τ=0.0000, and curvature prior σ=0.0°. Bundle-conditioned Markov chain absorbing tractography (MPAT) 1 was solved to tolerance 1.0e-6 (Minimal Residual (MR) acceleration). Up to 400 reactive-flux streamlines were integrated through the bundle's flux field with RK4 (step 0.35 voxels, 15–250 mm). A branch arbor with up to 6 cortical branches per bundle end was traced along the bundle's cross-section ridges.


  1. John Kruper, Matthew Cieslak, and Ariel Rokem. Probabilistic mapping of brain connections without random sampling. bioRxiv, 2026. URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766276v1, doi:10.64898/2026.11.12.766276. ↩↩