Skills histolab
🔬

histolab

Safe ⚡ Contains scripts⚙️ External commands

Process whole slide images for digital pathology

Also available from: davila7

Histolab automates tissue detection and tile extraction from gigapixel whole slide images. It processes WSI files to extract informative tiles for deep learning pipelines and medical research.

Supports: Claude Codex Code(CC)
🥉 72 Bronze
1

Download the skill ZIP

2

Upload in Claude

Go to Settings → Capabilities → Skills → Upload skill

3

Toggle on and start using

Test it

Using "histolab". Extract 100 tiles from prostate tissue slide

Expected outcome:

  • Loaded slide: prostate.svs (dimensions: 46000×32000 pixels)
  • Created RandomTiler with 512×512 tile size
  • Applied tissue mask filtering (80% threshold)
  • Extracted 100 tiles to output/prostate_tiles/
  • Generated preview visualization showing tile locations
  • Average tissue coverage: 87% across extracted tiles

Using "histolab". Create tissue mask and visualize

Expected outcome:

  • Initialized TissueMask with default filters
  • Generated binary mask (tissue: 72%, background: 28%)
  • Saved mask visualization to output/mask_preview.png
  • Detected 3 tissue regions with varying sizes

Security Audit

Safe
v4 • 1/17/2026

Documentation-only skill containing markdown files with Python code examples for histolab, a legitimate digital pathology library. All 389 static findings are false positives - backticks are markdown syntax for code blocks (not Ruby/shell execution), no actual cryptographic or malicious patterns exist, and no executable code is present. This is a safe scientific tool for processing whole slide images.

8
Files scanned
3,010
Lines analyzed
2
findings
4
Total audits

Risk Factors

⚡ Contains scripts (1)
SKILL.md:242
⚙️ External commands (1)
SKILL.md:17-19
Audited by: claude View Audit History →

Quality Score

45
Architecture
100
Maintainability
87
Content
22
Community
100
Security
96
Spec Compliance

What You Can Build

Prepare training datasets from pathology slides

Extract balanced tile datasets from whole slide images for training deep learning models in cancer detection and classification.

Build automated tissue analysis pipelines

Create reproducible workflows for tissue segmentation, tile extraction, and quality assessment across slide collections.

Standardize slide preprocessing workflows

Implement consistent tissue detection and tile extraction procedures for research studies and clinical trials.

Try These Prompts

Basic tile extraction 
Load the slide at path 'slide.svs' and extract 100 random tiles of size 512x512 pixels. Save them to 'output/tiles/' directory.
Tissue detection 
Create a tissue mask for my slide and visualize it. Use the BiggestTissueBoxMask to focus on the main tissue section.
Grid extraction workflow 
Extract tiles in a grid pattern across all tissue regions with 20% overlap. Use tissue mask to avoid background areas.
Quality-based selection 
Use ScoreTiler with NucleiScorer to extract the 50 tiles with highest nuclei density. Generate a report of tile scores.

Best Practices

  • Always preview tile locations with locate_tiles() before extraction to verify settings
  • Use appropriate pyramid level - level 0 for full resolution, level 1-2 for faster processing
  • Set tissue_percent threshold between 70-90% to balance coverage and quality

Avoid

  • Extracting all tiles at highest resolution without considering memory constraints
  • Using RandomTiler without setting a seed for reproducible results
  • Skipping tissue mask preview which may lead to extracting background tiles

Frequently Asked Questions

What file formats does histolab support?
Histolab supports common WSI formats including SVS, TIFF, NDPI, and other formats compatible with OpenSlide library.
How do I handle slides with multiple tissue sections?
Use TissueMask instead of BiggestTissueBoxMask to detect all tissue regions, or create custom masks for specific ROIs.
Can I extract tiles at multiple resolutions?
Yes, use the level parameter (0=full resolution, 1=half resolution, etc.) or extract hierarchically with multiple tilers.
How do I remove pen annotations from slides?
Create custom filters using HSV color space to detect and exclude pen markings based on their characteristic colors.
What is the difference between RandomTiler and GridTiler?
RandomTiler samples tiles randomly across tissue, good for training data. GridTiler extracts in systematic pattern for complete coverage.
How can I speed up extraction for large datasets?
Use lower pyramid levels, reduce tile count, use BiggestTissueBoxMask, and enable logging to monitor progress.

Developer Details

Author

K-Dense-AI

License

Apache-2.0 license

Repository

https://github.com/K-Dense-AI/claude-scientific-skills/tree/main/scientific-skills/histolab

Ref

main

File structure

📁 references/

📄 filters_preprocessing.md

📄 slide_management.md

📄 tile_extraction.md

📄 tissue_masks.md

📄 visualization.md

📄 evaluation_output.json

📄 SKILL.md