Tissue Classification

This repository contains a pipeline for tissue classification in whole-slide images (WSIs). It includes preprocessing utilities, dataset splitting tools, machine learning training and evaluation code, Hydra configurations, and submission scripts for running individual pipeline stages.

Repository Structure

• preprocessing/ contains source code for WSI preprocessing, tiling, mask generation, embedding extraction, quality control, and related statistics.
• ml/ contains machine learning code for data loading, model training, evaluation, callbacks, and prediction outputs.
• split/ contains utilities for train/test and k-fold dataset splitting.
• configs/ contains Hydra configuration files for preprocessing, splitting, logging, datasets, and machine learning experiments.
• scripts/ contains executable submission scripts for running pipeline stages.
• pyproject.toml and uv.lock define the Python project metadata and locked dependencies.
• LICENSE contains the MIT License for this repository.

Setup

The project uses Python 3.12 and uv for dependency management.

uv sync

Usage

Pipeline stages can be run through the corresponding scripts in scripts/. For example:

uv run python scripts/submit_tiling.py
uv run python scripts/submit_embeddings.py
uv run python scripts/submit_train_linear_probe.py

Configuration is managed with Hydra. Base configurations are stored in configs/, with task-specific configurations grouped by pipeline stage.