https://github.com/pabannier/fgfr3mut
Official Implementation of the "AI allows pre-screening of FGFR3 mutational status using routine histology slides of muscle-invasive bladder cancer" paper
https://github.com/pabannier/fgfr3mut
cancer deep-learning histopathology
Last synced: 4 months ago
JSON representation
Official Implementation of the "AI allows pre-screening of FGFR3 mutational status using routine histology slides of muscle-invasive bladder cancer" paper
- Host: GitHub
- URL: https://github.com/pabannier/fgfr3mut
- Owner: PABannier
- License: other
- Created: 2024-04-30T07:47:57.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-01-01T14:56:29.000Z (5 months ago)
- Last Synced: 2025-01-01T15:36:14.932Z (5 months ago)
- Topics: cancer, deep-learning, histopathology
- Language: Python
- Homepage: https://www.nature.com/articles/s41467-024-55331-6
- Size: 21.5 KB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# FGFR3MUT screening model
[Publication](https://www.nature.com/articles/s41467-024-55331-6)
This is an official implementation of the FGFR3MUT model. This model is intended for pre-screening tumors of muscle invasive bladder cancer (MIBC) harboring an FGFR3 mutation.
It provides the user with:
1. The model implementation (Chowder)
2. An inference script to replicate the experiments on TCGA
## Data Availability
We provide a script `download.py` which downloads the Chowder weights of the pre-screening model, as well as the slide features of `TCGA-BLCA` extracted with [Bioptimus' H0 extractor](https://github.com/bioptimus/releases/tree/main/models/h-optimus/v0?utm_source=owkin&utm_medium=referral&utm_campaign=h-bioptimus-o).
## Instructions
Start by installing the required dependencies:
```bash
# Clone the repo
git clone https://github.com/PABannier/fgfr3mut.git
cd fgfr3mut
# Create a virtual environment with all the requirements
conda create -n fgfr3mut_env python=3.8
conda activate fgfr3mut_env
# Install the `fgfr3mut` package
pip install .
```
Then, download the model weights and the TCGA features from [HuggingFace](https://huggingface.co/datasets/PABannier/fgfr3mut):
```shell
$ python download.py --out_dir
```
Note that the `TCGA-BLCA` slide features weigh more than 200GB. In comparison, the actual Chowder weights weigh 50MB.
This should take a few minutes depending on your network connection.
Now, you can run the experiments reported in the manuscript. The actual computations should take ten minutes to run on CPU and only a few seconds on GPU.
### External validation on TCGA-BLCA (MIBC cases only)
```shell
$ python fgfr3mut/run_inference.py \
--data_dir \
--device cpu \ # `cuda:0` if you have a GPU
--n_tiles 5000 \
--keep_tcga_cases MIBC
```
Inference on: n_slides=379, n_patients=308.
You should obtain the following results, reported in the publication:
| Metric | Value |
| ----------- | -------------------|
| AUC | 0.82 [0.75 - 0.88] |
| Sensitivity | 0.95 [0.92 - 1.00] |
| Specificity | 0.47 [0.30 - 0.72] |
| PPV | 0.14 [0.09 - 0.23] |
| NPV | 0.99 [0.98 - 1.00] |
### Comparison with Loeffler et al.
To run the external validation on the exact cases reported by [Loeffler et al.](), run:
```shell
$ python fgfr3mut/run_inference.py \
--data_dir \
--device cpu \ # `cuda:0` if you have a GPU
--n_tiles 5000 \
--loeffler_tcga_cases
```
Inference on: n_slides=391, n_patients=327.
You should obtain the following results, reported in the publication:
| Metric | Value |
| ----------- | ------------------ |
| AUC | 0.83 [0.77 - 0.88] |
| Sensitivity | 0.94 [0.92 - 0.98] |
| Specificity | 0.44 [0.24 - 0.58] |
| PPV | 0.23 [0.17 - 0.30] |
| NPV | 0.98 [0.96 - 0.99] |