Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/med-air/Endo-FM
[MICCAI'23] Foundation Model for Endoscopy Video Analysis via Large-scale Self-supervised Pre-train
https://github.com/med-air/Endo-FM
endoscopy foundation-model large-scale miccai2023 pre-train self-supervised video
Last synced: 3 months ago
JSON representation
[MICCAI'23] Foundation Model for Endoscopy Video Analysis via Large-scale Self-supervised Pre-train
- Host: GitHub
- URL: https://github.com/med-air/Endo-FM
- Owner: med-air
- License: apache-2.0
- Created: 2023-06-05T08:45:47.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-03-11T12:56:24.000Z (4 months ago)
- Last Synced: 2024-03-14T15:22:55.220Z (4 months ago)
- Topics: endoscopy, foundation-model, large-scale, miccai2023, pre-train, self-supervised, video
- Language: Python
- Homepage:
- Size: 38.2 MB
- Stars: 123
- Watchers: 2
- Forks: 11
- Open Issues: 6
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Lists
- Awesome-Healthcare-Foundation-Models - Endo-FM database
- Awesome-Reasoning-Foundation-Models - [code
README
# Foundation Model for Endoscopy Video Analysis
[//]: # (
)
[//]: # ( 
)
[//]: # (
)
[//]: # (---)
[//]: # ([](https://twitter.com/opendilab))
[//]: # ([](https://pypi.org/project/DI-engine/))
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ()
[//]: # ([](https://codecov.io/gh/opendilab/DI-engine))
[//]: # ()
[//]: # ()
[//]: # ([](https://github.com/Med-AIR/Endo-FM/stargazers))
[//]: # ([](https://github.com/Med-AIR/Endo-FM/network))
[//]: # ()
[//]: # ([](https://github.com/opendilab/DI-engine/issues))
[//]: # ([](https://github.com/opendilab/DI-engine/pulls))
[//]: # ([](https://github.com/opendilab/DI-engine/graphs/contributors))
[//]: # ([](https://github.com/Med-AIR/Endo-FM/blob/master/LICENSE))
[//]: # (Updated on 2023.06.09)
This repository provides the official PyTorch implementation of the paper [**Foundation Model for Endoscopy Video Analysis via Large-scale Self-supervised Pre-train**](https://arxiv.org/abs/2306.16741)
by [Zhao Wang](https://kyfafyd.wang)\*, [Chang Liu](https://scholar.google.com/citations?user=q2JSP3kAAAAJ)\*, [Shaoting Zhang](http://www.qingyuan.sjtu.edu.cn/a/Shaoting-Zhang.html)†, and [Qi Dou](http://www.cse.cuhk.edu.hk/~qdou)†.
## Key Features
[//]: # (key feature bulletin points here)
- First foundation model for endoscopy video analysis.
- A large-scale endoscopic video dataset with over 33K video clips.
- Support 3 types of downstream tasks, including classification, segmentation, and detection.
## Links
- [Paper](https://arxiv.org/abs/2306.16741)
- [Model](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EZh5mWE5CL1BpaJ1bXuokfYBDM2VaMknqG7YpaQBRgAvdQ?e=e2rVYW)
- [OpenMEDLab Page](https://github.com/openmedlab/Endo-FM)
## Details
> Foundation models have exhibited remarkable success in various applications, such as disease diagnosis and text report generation. To date, a foundation model for endoscopic video analysis is still lacking. In this paper, we propose Endo-FM, a foundation model specifically developed using massive endoscopic video data. First, we build a video transformer, which captures both local and global long-range dependencies across spatial and temporal dimensions. Second, we pre-train our transformer model using global and local views via a self-supervised manner, aiming to make it robust to spatial-temporal variations and discriminative across different scenes. To develop the foundation model, we construct a large-scale endoscopy video dataset by combining 9 publicly available datasets and a privately collected dataset from Baoshan Branch of Renji Hospital in Shanghai, China. Our dataset overall consists of over 33K video clips with up to 5 million frames, encompassing various protocols, target organs, and disease types. Our pre-trained Endo-FM can be easily adopted for a given downtream task via fine-tuning by serving as the backbone. With experiments on 3 different types of downstream tasks, including classification, segmentation, and detection, our Endo-FM surpasses the current state-of-the-art self-supervised pre-training and adapter-based transfer learning methods by a significant margin.
[//]: # (More intro text here.)
## Datasets
We utilize 6 public and 1 private datasets for pre-training and 3 datasets as the downstream tasks.
Except for SUN & SUN-SEG, we provide our preprocessed data for pre-training and downstream tasks.
#### Pre-training Data (6 public + 1 private)
- Colonoscopic [[original paper]](https://ieeexplore.ieee.org/abstract/document/7442848) [[original dataset]](http://www.depeca.uah.es/colonoscopy_dataset/) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/ES_hCHb2XWFJgsK4hrKUnNUBx3fl6QI3yyk9ImP4AkkRVw?e=LC4DU5)
- SUN & SUN-SEG [[original paper1]](https://www.sciencedirect.com/science/article/pii/S0016510720346551) [[original paper2]](https://link.springer.com/article/10.1007/s11633-022-1371-y) [[original dataset1]](http://amed8k.sundatabase.org/) [[original dataset2]](https://github.com/GewelsJI/VPS/blob/main/docs/DATA_PREPARATION.md)
- LPPolypVideo [[original paper]](https://link.springer.com/chapter/10.1007/978-3-030-87240-3_37) [[original dataset]](https://github.com/dashishi/LDPolypVideo-Benchmark) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/ERTYntGNWfZKj8FVjzsK0QEB6W6KoiuiP89Y3on1PJBAmg?e=P24jjG)
- Hyper-Kvasir [[original paper]](https://www.nature.com/articles/s41597-020-00622-y) [[original dataset]](https://datasets.simula.no/hyper-kvasir/) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EeHnnUmGbmBGlw7UlNVvw2wBzBMzKi8Sus5LrdwrQi-XUA?e=gWr5qH)
- Kvasir-Capsule [[original paper]](https://www.nature.com/articles/s41597-021-00920-z) [[original dataset]](https://datasets.simula.no/kvasir-capsule/) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EQhyk3_yz5pAtdpKVFU93S0BfPfTNpblPFXTHaW-BIjV-Q?e=9duP5z)
- CholecTriplet [[original paper]](https://www.sciencedirect.com/science/article/pii/S1361841522000846) [[original dataset]](https://cholectriplet2021.grand-challenge.org/) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/Ea6g5KpHaJNLvYFqoZpHeroBS801guoB16X18F4GfEG4pw?e=SWHoyQ)
- Our Private [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EZ2Vs0zU-L1Go8RITgs42b4BjlWy6UtGXh6AHmBGD_gGFw?e=SRiD7m)
#### Downstream Data (3 public)
- PolypDiag [[original paper]](https://link.springer.com/chapter/10.1007/978-3-031-16437-8_9) [[original dataset]](https://github.com/tianyu0207/weakly-polyp) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/Ed_RCZ86IktKkGNNL5qX9IsBvNa7wcyM8q4yBQBkzaBj8g?e=pvuZVt)
- CVC-12k [[original paper]](https://www.sciencedirect.com/science/article/pii/S0895611115000567) [[original dataset]](https://polyp.grand-challenge.org/Databases/) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EQzj78YsrVZAtbNVHW7WPEEBX1AeolLI7gmBkg-iEg1lQg?e=0gQPzy)
- KUMC [[original paper]](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0255809) [[original dataset]](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/FCBUOR) [[our preprocessed dataset]](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EQHKl1-MgA5Ams_sQ4_ssg8BFyd66qucAxUTEHz4lHxE7g?e=fFtXzd)
For SUN & SUN-SEG, you need first request the original videos following [this instruction](https://github.com/GewelsJI/VPS/blob/main/docs/DATA_PREPARATION.md).
Then, you can transfer the data for pre-training videos by the following:
```bash
cd Endo-FM/data
python sun.py
python sun_seg.py
python trans_videos_pretrain.py
```
Finally, generating the video list `pretrain/train.csv` for pre-training by the following:
```bash
cd Endo-FM/data
python gencsv.py
```
## Get Started
#### Main Requirements
- torch==1.8.0
- torchvision==0.9.0
- pillow==6.2.2
- timm==0.4.12
#### Installation
We suggest using Anaconda to setup environment on Linux, if you have installed anaconda, you can skip this step.
```shell
wget https://repo.anaconda.com/archive/Anaconda3-2020.11-Linux-x86_64.sh && zsh Anaconda3-2020.11-Linux-x86_64.sh
```
Then, we can install packages using provided `environment.yaml`.
```shell
cd Endo-FM
conda env create -f environment.yaml
conda activate endofm
```
#### Pre-trained Weights
You can directly download our pre-trained Endo-FM via this [link](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EZh5mWE5CL1BpaJ1bXuokfYBDM2VaMknqG7YpaQBRgAvdQ?e=e2rVYW) and put it under `checkpoints/` for downstream fine-tuning.
#### Downstream Fine-tuned Weights
Also, we provide the pre-trained weights of 3 downstream tasks for direct downstream testing.
| Dataset | PolypDiag | CVC-12k | KUMC |
|:--------------:|:----:|:----:|:-----:|
| Our Paper | 90.7 | 73.9 | 84.1 |
| Released Model | 91.5 | 76.6 | 84.0 |
| Weights | [link](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/ERSlUP10MGpBuhg1uN5iaHABKqz1SPQSrr03j4sEWey-bw?e=muv8RL) | [link](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EePnpTllUCFEqpYp6BFPv0sBQyST4CV4jQ8pvaRynCkD7Q?e=f7LeBx) | [link](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155167044_link_cuhk_edu_hk/EYPkwbFyMfxEirezWtumAGIBSCTQ0EvDN4u99KKiRsaVBA?e=DsrkVG) |
#### Pre-training
```shell
cd Endo-FM
wget -P checkpoints/ https://github.com/kahnchana/svt/releases/download/v1.0/kinetics400_vitb_ssl.pth
bash scripts/train_clips32k.sh
```
#### Downstream Fine-tuning
```shell
# PolypDiag (Classification)
cd Endo-FM
bash scripts/eval_finetune_polypdiag.sh
# CVC (Segmentation)
cd Endo-FM/TransUNet
python train.py
# KUMC (Detection)
cd Endo-FM/STMT
python setup.py build develop
python -m torch.distributed.launch \
--nproc_per_node=1 \
tools/train_net.py \
--master_port=$((RANDOM + 10000)) \
--config-file configs/STFT/kumc_R_50_STFT.yaml \
OUTPUT_DIR log_dir/kumc_finetune
```
#### Direct Downstream Testing
```shell
# PolypDiag (Classification)
cd Endo-FM
bash scripts/test_finetune_polypdiag.sh
# CVC (Segmentation)
cd Endo-FM/TransUNet
python train.py --test
# KUMC (Detection)
cd Endo-FM/STMT
python setup.py build develop
python -m torch.distributed.launch \
--nproc_per_node=1 \
tools/test_net.py \
--master_port=$((RANDOM + 10000)) \
--config-file configs/STFT/kumc_R_50_STFT.yaml \
MODEL.WEIGHT kumc.pth \
OUTPUT_DIR log_dir/kumc_finetune
```
## 🙋♀️ Feedback and Contact
For further questions, pls feel free to contact [Zhao Wang](mailto:[email protected]).
## 🛡️ License
This project is under the Apache License 2.0 license. See [LICENSE](LICENSE) for details.
## 🙏 Acknowledgement
Our code is based on [DINO](https://github.com/facebookresearch/dino), [TimeSformer](https://github.com/facebookresearch/TimeSformer), [SVT](https://github.com/kahnchana/svt), [TransUNet](https://github.com/Beckschen/TransUNet), and [STFT](https://github.com/lingyunwu14/STFT). Thanks them for releasing their codes.
## 📝 Citation
If you find this code useful, please cite in your research papers.
```
@inproceedings{
wang2023foundation,
title={Foundation Model for Endoscopy Video Analysis via Large-scale Self-supervised Pre-train},
author={Zhao Wang and Chang Liu and Shaoting Zhang and Qi Dou},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={101--111},
year={2023},
organization={Springer}
}
```