Data Preparation

• MIMIC-CXR-JPG: https://www.physionet.org/content/mimic-cxr-jpg/2.1.0/

• CheXpert: https://www.kaggle.com/datasets/ashery/chexpert

• NIH: https://www.kaggle.com/organizations/nih-chest-xrays/datasets

• VinDr-CXR: https://physionet.org/content/vindr-cxr/

• MedFMC: https://github.com/openmedlab/MedFM

• RSNA: https://www.kaggle.com/c/rsna-pneumonia-detection-challenge

• SIIM：https://academictorrents.com/details/6ef7c6d039e85152c4d0f31d83fa70edc4aba088

In this study, we clean the pre-training dataset by filtering out corrupted images and removing large black borders from the images.

Once downloaded the dataset, please change the data roots in data_utils/data_path.py.