Atlases & masks
atlas_mask_tools/ holds atlas and mask generation utilities together with ready-made brain
templates for the regions we return to repeatedly.
Most of what you need is already in CANlab.
Neuroimaging_Pattern_Masks is a large,
curated collection of parcellations, region masks and published multivariate signatures, and
CanlabCore’s atlas object provides the methods for working with them — selecting regions,
combining atlases, extracting data. The
CANlab atlas documentation is the reference. Build a custom
mask only when nothing suitable exists.
What we build on top
The tooling here covers the cases our work needs that a general atlas does not supply:
- Study-specific ROI masks, particularly for brainstem and hypothalamic regions central to gut-brain signalling — the nucleus of the solitary tract, for example, which is small enough that atlas choice materially affects whether an effect is detectable
- Combining or restricting atlases — intersecting a parcellation with a group-level mask so that regions with poor coverage do not silently contribute
- Templates for repeated use across studies, so the same anatomical definition is applied consistently rather than redrawn per project
Coverage matters here
Ventral brain regions suffer signal dropout in EPI, and those are exactly the regions our research lines care about. Two practical consequences:
- Check the group coverage mask before interpreting any ROI result. A region can appear null simply because half the subjects have no usable signal there.
- Restrict analyses to voxels with adequate coverage across subjects, and report the restriction.
Signatures
Published multivariate signatures — NPS, SIIPS, PINES and others in Neuroimaging_Pattern_Masks — are applied as a priori measures during second-level analysis. Applying an established signature is a far stronger test than an exploratory whole-brain search, because the pattern and its interpretation were fixed before your data existed.
The batch pipeline produces a dedicated signature response report. See the CANlab batch documentation.