---
title: Pathology on the platform
description: Digital pathology analysis for whole slide images — tiling, tissue detection, and deep-learning feature extraction.
tags:
  - pathology
---

Analyze histology and whole slide images computationally — open large WSI formats, cut them into quality-filtered tiles, detect and measure tissue regions, and extract deep-learning features for downstream analysis. Reach for these tools when your data is digital pathology imagery and you want tile-level and tissue-level readouts at the slide scale. They sit in the tissue-scale, translational analysis arc rather than the small-molecule chemistry pipeline. See the [Cell & Tissue overview](/docs/cell-tissue) for where this fits.

## What pathology analysis does

You give the tool a whole slide image stored in the platform's workspace and pick a **workflow**. The processor opens the slide (via histolab/OpenSlide, falling back to PIL for standard image formats) and runs the workflow you chose over the pixels, returning structured results plus downloadable artifacts.

The three operational workflows are:

- **Tile extraction** — cut the slide into a grid of tiles, filter out background and blurry tiles, and package the survivors as a downloadable archive. This is the default and the foundation for most tile-based pipelines.
- **Tissue detection** — segment tissue from background across the slide, compute the tissue-covered percentage, and return bounding-box regions plus a mask image.
- **Feature extraction** — run each retained tile through a pretrained ResNet-50 and emit a per-tile deep-learning feature matrix for clustering, classification, or similarity work.

The schema also lists **classification** and **cell segmentation**, but the backend does not implement them: a job that requests either fails with an "Unsupported workflow" error. For tissue typing or cell counts, export the tiles or feature matrix and run your own classifier.

## Citations

> Goode A, Gilbert B, Harkes J, Jukic D, Satyanarayanan M. OpenSlide: A vendor-neutral software foundation for digital pathology. *Journal of Pathology Informatics*. 2013;4:27. doi: [10.4103/2153-3539.119005](https://doi.org/10.4103/2153-3539.119005)

> histolab (open-source software, no reference publication; the platform installs `histolab>=0.6.0`). [https://github.com/histolab/histolab](https://github.com/histolab/histolab)

> He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recognition. *2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)*. 2016:770-778. doi: [10.1109/CVPR.2016.90](https://doi.org/10.1109/CVPR.2016.90)

> Reinhard E, Adhikhmin M, Gooch B, Shirley P. Color transfer between images. *IEEE Computer Graphics and Applications*. 2001;21(4):34-41. doi: [10.1109/38.946629](https://doi.org/10.1109/38.946629)

> Macenko M, Niethammer M, Marron JS, Borland D, Woosley JT, Guan X, Schmitt C, Thomas NE. A method for normalizing histology slides for quantitative analysis. *2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro*. 2009:1107-1110. doi: [10.1109/ISBI.2009.5193250](https://doi.org/10.1109/ISBI.2009.5193250)

> Vahadane A, Peng T, Sethi A, Albarqouni S, Wang L, Baust M, et al. Structure-Preserving Color Normalization and Sparse Stain Separation for Histological Images. *IEEE Transactions on Medical Imaging*. 2016;35(8):1962-1971. doi: [10.1109/TMI.2016.2529665](https://doi.org/10.1109/TMI.2016.2529665)

## When to reach for it

- **You explicitly have whole slide images** (SVS, NDPI, TIFF) and want tiling or tissue detection.
- **You mention histopathology or digital pathology image analysis** and need cell- or tissue-level readouts.

Reach for a different tool when you want molecular analysis (use the omics tools) or a general drug-discovery / target-identification workflow — pathology analysis is specifically for slide imagery.

## Key capabilities

- **WSI processing**: SVS, TIFF, NDPI, and standard image formats (PNG/JPG)
- **Tile extraction**: grid tiling with tissue-percentage and blur quality filters
- **Tissue detection**: Otsu segmentation, tissue-coverage percentage, and region bounding boxes
- **Feature extraction**: pretrained ResNet-50 features per tile (2048-dim)
- **Stain normalization**: reinhard, macenko, or vahadane

 Large WSI files are processed in tiles — ensure sufficient storage for intermediate results.

## How it fits the platform

Pathology analysis runs on GPU (A10G), asynchronously, and can take **10–60 minutes** depending on slide size and workflow. It reads the slide from workspace storage and writes its artifacts — tile archives, masks, or feature files — back for you to download or feed into downstream analysis. Results render on the canvas as an intelligence card.

## Read next