insula-rtop: a pure-Python reproduction of the diffusion analyses of Menon et al., eLife 2020;9:e53470

A pure-Python re-implementation of the diffusion-MRI analyses behind Figures 2, 3, 4, 7 and 8 of Menon et al., Microstructural organization of human insula is linked to its macrofunctional circuitry and predicts cognitive control, eLife 2020;9:e53470: RTOP from MAP-MRI, projection to the fs_LR cortical surface, insula and ACC subdivisions, ANOVA and subsampling stability, gradient directional statistics, and canonical correlation with cognitive control.

uv sync is the whole installation. There is no Connectome Workbench, FSL or FreeSurfer dependency: the HCP structural package already ships midthickness surfaces in the space the diffusion data lives in, so Appendix 1's two-step projection collapses to one trilinear interpolation, and every other binary is replaced by the closed-form operation it performed. 254 tests pin the numerical core to analytic answers rather than to remembered output.

Reproduced over 451 HCP Young Adult participants. The subdivision ordering vAI < dAI < PI holds in both hemispheres under both parcellations, the directional statistics reproduce the published confidence intervals, and the brain-behaviour prediction reproduces at r = 0.155 out of sample against 0.19.

Three findings are not in the paper. The brain-behaviour prediction does not survive adjustment for sex. The left-hemisphere ACC ordering turns on where a single ROI boundary falls, and moving it 7 mm reproduces the published ordering in both hemispheres. The Figure 3C mean directions reproduce to 9 degrees once angles are read in the anatomical plane the figure is labelled with. All three are documented in the README with the evidence and the commands to re-run them.

No Human Connectome Project data is redistributed. Reproducing the results requires your own HCP Young Adult access.