You need to download the images of each dataset from their original sources. We recommend downloading the images to the data/datasets/ directory. If you choose to download the images to a different directory, please make sure to update the path in the scripts accordingly. Please refer to the following instructions for more details.
-
MIMIC-CXR: Download the images from https://physionet.org/content/mimic-cxr-jpg/2.1.0/. We also provided the metadata we organized in
data/datasets/MIMIC-CXR/MIMIC-CXR_metadata.json, including the finds extracted from the radiology reports using GPT-4. -
NIH-CXR: Download the images from https://www.kaggle.com/datasets/nih-chest-xrays/data. An example image path in our split is
NIH-CXR/images_002/images/00001906_004.png. -
CheXpert: Download the images from https://stanfordaimi.azurewebsites.net/datasets/8cbd9ed4-2eb9-4565-affc-111cf4f7ebe2/. An example image path in our split is
CheXpert-v1.0/train/patient10900/study25/view1_frontal.jpg. -
Pneumonia: Download the images from https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia. An example image path in our split is
pneumonia/images/test/PNEUMONIA/person112_bacteria_538.jpeg. -
COVID-QU: Download the images from https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database. An example image path in our split is
COVID-QU/images/COVID-3330.png. -
Open-i: Download the images from https://www.kaggle.com/datasets/raddar/chest-xrays-indiana-university. An example image path in our split is
open-i/images/1907_IM-0589-1001.dcm.png. -
VinDr-CXR: Download the images from https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/data. An example image path in our split is
vindr-cxr/images/e1eb9553f694d0eba82535625d70186c.png.
-
ISIC: Download the entire archive (81621 images) using official command line tool https://github.com/ImageMarkup/isic-cli?tab=readme-ov-file#downloading-images. An example image path in our split is
ISIC/images/ISIC_0060248.JPG. We also provided the metadata we organized indata/datasets/ISIC/ISIC_metadata.json, including the radiology reports generated by GPT-4V. -
HAM10000: Download the images from https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/DBW86T. An example image path in our split is
HAM10000/images/ISIC_0033762.jpg. -
PAD-UFES-20: Download the images from https://www.kaggle.com/datasets/mahdavi1202/skin-cancer. An example image path in our split is
PAD-UFES-20/images/PAT_1850_3631_519.png. -
Melanoma: Download the images from https://www.kaggle.com/datasets/hasnainjaved/melanoma-skin-cancer-dataset-of-10000-images/data. An example image path in our split is
Melanoma/test/benign/melanoma_9999.jpg. -
UWaterloo: Download the images from https://uwaterloo.ca/vision-image-processing-lab/research-demos/skin-cancer-detection. An example image path in our split is
UWaterloo/notmelanoma/dermquest/D39_orig.jpg.