# Tile and patch encoding of slides

Source: https://onco.cc/terms/tile-patch-encoding/  
OnCo record `tile-patch-encoding` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

A whole slide is cut into thousands of small square tiles, each tile is turned into a vector by an image model, and the vectors are pooled to describe the slide.

## Summary

Multiple-instance learning is supervised learning in which the learner receives labelled bags of instances rather than individually labelled instances (Wikipedia); a slide is the bag and its tiles the instances, since the diagnosis is known for the slide but not for each square. Tiles (typically 224 to 256 pixels at 20x) are encoded by a frozen pathology foundation model such as UNI or Virchow, and an attention-based pooling (ABMIL, CLAM) weights the tiles to make a slide-level prediction while showing which regions drove it.

## Fields

- Kind: Term
- Last checked: 2026-09-24
- Also known as: tile encoding; patch encoding; tile embeddings; patch embeddings; tiling; slide tiles; slide patches; tile-level features
- Tags: cansim-terms

## Notes

- Listed in the CanSim terms map 1.0.0 (docs/onco/terms.json, generated 2026-09-24), CC BY 4.0, attribution: CanSim project, an open, public-data-first cancer foundation-model programme; CanSim page path /terms/tile-patch-encoding.

## Sources

- Wikipedia: https://en.wikipedia.org/wiki/Multiple_instance_learning
- Lu et al., CLAM: data-efficient and weakly supervised computational pathology on whole-slide images (Nature Biomedical Engineering 2021): https://doi.org/10.1038/s41551-020-00682-w
- Wikipedia: https://en.wikipedia.org/wiki/Multiple_instance_learning

## Connected records

- terms: [Attention-based multiple-instance learning (ABMIL, CLAM)](https://onco.cc/terms/abmil/), [Cancer AI vocabulary (CanSim terms map)](https://onco.cc/terms/cancer-ai-vocabulary/), [Digital pathology and whole-slide images (WSI)](https://onco.cc/terms/digital-pathology-wsi/), [H&E staining (haematoxylin and eosin)](https://onco.cc/terms/h-and-e-staining/), [Magnification (20x, 40x) and microns per pixel](https://onco.cc/terms/magnification/), [Pathology foundation models: UNI, UNI2, Virchow2, CTransPath, CONCH, TITAN](https://onco.cc/terms/pathology-foundation-models/)
- technologies: [Pathology & radiology foundation models](https://onco.cc/technologies/pathology-foundation-model/)

---
JSON: https://onco.cc/api/v1/entities/tile-patch-encoding.json