Home

Awesome

RadFM

The official code for the paper "Towards Generalist Foundation Model for Radiology by Leveraging Web-scale 2D&3D Medical Data"

ArXiv

Website

Model checkpoint

In this project, we collect a large-scale medical multi-modal dataset, MedMD, with 16M 2D or 3D images. We train a new medical multi-modal generative model RadFM on it, enabling both 2D and 3D scans, multi-image input and visual-language interleaving cases.

<img src="https://github.com/chaoyi-wu/RadFM/blob/main/Images/GIF.gif"/>

Latest News:

All Datasets are released! We have updated the links in our dataset table. You can find all our text part data in https://huggingface.co/datasets/chaoyi-wu/RadFM_data_csv.

For decompressing the splited compression files in most cases, please check the following code in linux:

cat zip.z* > myzip.zip
unzip myzip.zip

Quick Start:

For quick start, you can check the Quick_demo path.
We demonstrate a simple diagnosis case here to show how to inference with our model.
Feel free to modify it as you want.

By the way, never try to perform this in cpu and gpu is all you need :).

Pre-train:

For re-training a model on our dataset or large-scale testing our pre-train model, you can check src.

Simply, train.py for training and test.py for testing.

Case Study:

Some cases produced by our final model:

<img src="https://github.com/chaoyi-wu/RadFM/blob/main/Images/result_vqa.jpg"/> <img src="https://github.com/chaoyi-wu/RadFM/blob/main/Images/result_report.jpg"/> <img src="https://github.com/chaoyi-wu/RadFM/blob/main/Images/result_rationale.jpg"/>

Dataset-Links:

MedKD Dataset downloading URL:

Dataset NameLinkAccess
Rad3D-series-Closed
MPx-series-Closed
PMC-Figureshttps://pan.baidu.com/s/1Src_rhXsaOFp8zJ_3zMFsQ?pwd=p3neOpen Access
PMC-Inlinehttps://huggingface.co/datasets/chaoyi-wu/PMC-InlineOpen Access
PMC-CaseReportOriginal version, Filtered versionOpen Access
VinDr-Mammohttps://www.physionet.org/content/vindr-mammo/1.0.0/Credentialed Access
VinDr-SpineXRhttps://www.physionet.org/content/vindr-spinexr/1.0.0/Credentialed Access
VinDr-PCXRhttps://physionet.org/content/vindr-pcxr/1.0.0/Credentialed Access
PMC-OAhttps://huggingface.co/datasets/axiong/pmc_oa_betaOpen Access
PMC-VQAhttps://huggingface.co/datasets/xmcmic/PMC-VQAOpen Access
VQA-RADhttps://osf.io/89kps/Open Access
SLAKEhttps://www.med-vqa.com/slake/Open Access
MIMIC-CXRhttps://physionet.org/content/mimic-cxr/2.0.0Credentialed Access
VinDr-CXRhttps://physionet.org/content/vindr-cxr/1.0.0/Credentialed Access
NIH ChestXray14https://nihcc.app.box.com/v/ChestXray-NIHCC/folder/36938765345Open Access
CheXperthttps://aimi.stanford.edu/chexpert-chest-x-raysOpen Access
Covid-CXR2https://www.kaggle.com/datasets/andyczhao/covidx-cxr2Open Access
NLM-TBMontgomery, ChinaSetOpen Access
Object-CXRhttps://web.archive.org/web/20201127235812/https://jfhealthcare.github.io/object-CXR/Open Access
OpenIhttps://www.kaggle.com/datasets/raddar/chest-xrays-indiana-universityOpen Access
RSNAhttps://www.rsna.org/education/ai-resources-and-training/ai-image-challenge/rsna-pneumonia-detection-challenge-2018Open Access
SIIM-ACRhttps://www.kaggle.com/datasets/jesperdramsch/siim-acr-pneumothorax-segmentation-dataOpen Access

Acknowledgment:

We sincerely thank all the contributors who uploaded the relevant data in our dataset online. We appreciate their willingness to make these valuable cases publicly available.

Contact

If you have any questions, please feel free to contact wtzxxxwcy02@sjtu.edu.cn.