---

name: segmentation-labels description: "Use ANTsPy segmentation and label workflows for Atropos, k-means, Otsu, priors, joint label fusion, label statistics, overlap, geometry, matrices, centroids, points, and multi-label morphology." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

ANTsPy Segmentation and Labels

Use this sub-skill when a task asks for ANTsPy segmentation, label images, connected components, label statistics, overlap/geometry measures, label matrices, centroid or point-image workflows, multi-label morphology, or segmentation-to-mask/registration handoffs.

Start Here

1. Import the public package as import ants; the package distribution is antspyx.
2. Check verified signatures and return contracts in API reference.
3. Pick a recipe from workflows: k-means, Atropos, Otsu, prior-based segmentation, joint label fusion, label stats/overlap/matrix, point images, or mask alignment validation.
4. Use troubleshooting before changing class counts, masks, priors, label dtypes, or physical metadata.
5. Run scripts/antspy_segmentation_smoke.py for a tiny deterministic in-memory check in an environment with antspyx installed.

Core Contracts

• ants.atropos(...), ants.kmeans_segmentation(...), and ants.prior_based_segmentation(...) return dictionaries with segmentation and probabilityimages; ants.otsu_segmentation(...) returns only an ANTsImage label image.
• ants.joint_label_fusion(...) returns an ANTsImage for intensity-only fusion when no labels are supplied, and a dictionary with segmentation, intensity, probabilityimages, and segmentation_numbers when label_list is supplied.
• ants.fuzzy_spatial_cmeans_segmentation(...) and ants.functional_lung_segmentation(...) use snake_case keys segmentation_image and probability_images; lung segmentation also returns processed_image.
• Label-measure helpers generally clone labels to unsigned int; non-integer or out-of-range labels can fail or silently collapse if not validated first.
• Overlap, statistics, matrices, and label/image interactions assume matching dimension, shape, origin, spacing, and direction; validate with image-core helpers before computing metrics.
• Register intensity images with registration-transforms, apply transforms to labels with nearest-neighbor or label-safe interpolation, then return here for label statistics and overlap.

Route Elsewhere

• Image creation, IO, cloning, metadata repair, dtype conversion, image_physical_space_consistency, and coordinate transforms: image-core.
• Generic masks, thresholding, binary/grayscale morphology, smoothing, resampling, bias correction, and iMath: image-ops-math.
• Registration mechanics, transform lists, and label-safe transform application: registration-transforms.
• Plotting label overlays or exporting images for visualization: visualization-interop.
• Learning one-hot segmentation, patch extraction, and augmentation workflows: learning-deeplearn.

References

• API reference: verified public signatures, return keys, label dataframe contracts, dtype rules, and expensive-function caveats.
• Workflows: self-contained recipes for segmentation, priors, labels, overlap/statistics, label matrices, centroids, points, and registration/mask interactions.
• Troubleshooting: Atropos probability failures, mask/image mismatch, class ordering, label dtype handling, joint label fusion/KellyKapowski costs, and physical-space consistency before overlap.