---

name: image-preprocessing description: "Use TIAToolbox for tissue masking, stain normalization/extraction/augmentation, patch extraction, tile pyramids, and preprocessing validation before inference." disable-model-invocation: true metadata: disco-role: operating license: NOASSERTION

Image Preprocessing

Use this sub-skill when a task needs to prepare pathology images or WSIs before analysis: build tissue masks, normalize stain appearance, augment stain variation, extract fixed or point-centered patches, generate tile pyramids, or validate preprocessing choices before model inference.

Route Here For

• Tissue mask generation with OtsuTissueMasker, MorphologicalMasker, or the tissue-mask command.
• Stain normalization with get_normalizer(...), custom stain matrices, target images, or the stain-norm command.
• Stain extraction and augmentation with RuifrokExtractor, MacenkoExtractor, VahadaneExtractor, CustomExtractor, and StainAugmentor.
• Patch extraction from arrays, image files, WSI readers, masks, annotation stores, points, or sliding windows.
• Tile pyramid generation with TilePyramidGenerator, ZoomifyGenerator, and AnnotationTileGenerator.
• Lightweight sanity checks for preprocessing parameters before starting expensive model inference.

Route Elsewhere

• Use wsi-io for WSI opening, metadata, slide dimensions, thumbnails, read_rect, read_bounds, and coordinate-space questions.
• Use model-inference for patch prediction, semantic segmentation, nucleus detection, feature extraction, and inference engine batching.
• Use cli-and-configuration for complete CLI syntax, shared Click options, file discovery behavior, logging flags, and command orchestration.

References

• references/api-reference.md summarizes APIs, required arguments, defaults, and decision points.
• references/workflows.md gives copyable recipes for masks, stain workflows, patches, pyramids, and preprocessing validation.
• references/troubleshooting.md maps common preprocessing failures to concrete checks and fixes.
• scripts/preprocessing_smoke.py provides tiny in-memory checks for importability, maskers, patch coordinates/filtering, stain normalizer construction, and stain augmentor construction.

Fast Start

import numpy as np
from tiatoolbox.tools.tissuemask import MorphologicalMasker
from tiatoolbox.tools.stainnorm import get_normalizer
from tiatoolbox.tools.patchextraction import get_patch_extractor

thumbnail = np.zeros((1, 128, 128, 3), dtype=np.uint8)
thumbnail[:, 24:96, 24:96] = [120, 45, 90]
mask = MorphologicalMasker(kernel_size=3).fit_transform(thumbnail)[0]

normalizer = get_normalizer("reinhard")
normalizer.fit(thumbnail[0])
normalized = normalizer.transform(thumbnail[0])

extractor = get_patch_extractor(
    "slidingwindow",
    input_img=normalized,
    patch_size=(32, 32),
    stride=(32, 32),
    input_mask=mask,
    min_mask_ratio=0.1,
)
patches = list(extractor)

Before running a model, confirm that image and mask resolutions align, patch_size and stride are intentional, stain normalization has a fitted target, and patch counts are plausible for the tissue area.