DigitalSlideArchive/HistomicsTK

HistomicsTK – Open‑source toolkit for digital pathology image analysis

What it is – A Python library (and optional plugin for the Digital Slide Archive) that provides ready‑to‑use algorithms for working with whole‑slide and microscopy images. It focuses on tasks that are common in computational pathology such as color normalization, stain deconvolution, nuclei segmentation, and feature extraction.

Why it matters – Whole‑slide imaging produces gigapixel images that are difficult to handle with generic image‑processing tools. HistomicsTK wraps the large_image library for efficient reading of these formats and adds domain‑specific methods, enabling researchers and pathologists to apply machine‑learning pipelines to large histology datasets without building low‑level code from scratch.

How you use it

• Pure Python – Install via pip (optionally pulling pre‑built binary wheels for large_image) and call functions directly on image files or NumPy arrays. The API docs and example notebooks show end‑to‑end workflows.
• Within the Digital Slide Archive (DSA) – Deploy the library as a plugin so that users can launch analysis jobs from the web‑based HistomicsUI interface. Custom algorithms can be wrapped as slicer‑cli‑web containers and made available to the DSA platform.

Key capabilities

• Color handling – Normalization across staining batches and deconvolution to separate hematoxylin/eosin components.
• Segmentation – Classical nuclei segmentation pipelines (thresholding, watershed, morphological post‑processing).
• Feature extraction – Compute morphology, texture, and intensity descriptors for each segmented object.
• Annotation integration – Read and write annotation data used by HistomicsUI, enabling interactive review of algorithm outputs.

Installation snapshot

# Linux – from PyPI with pre‑built large_image wheels
python -m pip install histomicstk --find-links https://girder.github.io/large_image_wheels

# From source (Linux/macOS) – editable install
git clone https://github.com/DigitalSlideArchive/HistomicsTK.git
cd HistomicsTK
python -m pip install -e . --find-links https://girder.github.io/large_image_wheels

Windows requires additional steps (installing large-image, CMake, Visual Studio C++ tools, and the large-image-source-tiff package) as detailed in the README.

Where to learn more

• API reference: https://digitalslidearchive.github.io/HistomicsTK/api-docs.html
• Example notebooks: https://digitalslidearchive.github.io/HistomicsTK/examples.html
• Project website & demos: https://digitalslidearchive.github.io/digital_slide_archive/demos-examples/
• Source code for related components: Digital Slide Archive, HistomicsUI, large_image, slicer_cli_web.

Funding & community – Supported by NIH grant U24‑CA194362‑01. Development is coordinated through GitHub discussions and the broader Digital Slide Archive ecosystem.