The official codes for Prompt-driven Healthy/Diseased Image Pairs Enabling Pixel-level Chest X-ray Pathology Localization.
To clone all files:
git clone git@github.com:kaimingd/PIXEL.git
To install Python dependencies:
conda create -n pixel python=3.8
conda activate pixel
pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113
pip install -r requirements.txt
Note that the complete data file and model training logs/checkpoints can be download from link : https://pan.baidu.com/s/1oPoYeFsia3ngIsrSEWgX_w?pwd=refg (refg).
1. MIMIC-CXR Dataset
Navigate to MIMIC-CXR Database to download the training dataset. Note: in order to gain access to the data, you must be a credentialed user as defined on PhysioNet.
2. CheXpert Dataset
The CheXpert dataset consists of chest radiographic examinations from Stanford Hospital, performed between October 2002 and July 2017 in both inpatient and outpatient centers. Population-level characteristics are unavailable for the CheXpert test dataset, as they are used for official evaluation on the CheXpert leaderboard.
The main data (CheXpert data) supporting the results of this study are available at https://aimi.stanford.edu/chexpert-chest-x-rays.
1. CheXpert Dataset
The CheXpert test dataset has recently been made public, and can be found by following the steps in the cheXpert-test-set-labels repository.
2. PadChest Dataset
The PadChest dataset contains chest X-rays that were interpreted by 18 radiologists at the Hospital Universitario de San Juan, Alicante, Spain, from January 2009 to December 2017. The dataset contains 109,931 image studies and 168,861 images. PadChest also contains 206,222 study reports.
The PadChest is publicly available at https://bimcv.cipf.es/bimcv-projects/padchest. Those who would like to use PadChest for experimentation should request access to PadChest at the link.
3. ChestX-Det-10 dataset
ChestX-Det10 is a subset with instance-level box annotations of NIH ChestX-14. For image downloading, please visit http://resource.deepwise.com/xraychallenge/train_data.zip and http://resource.deepwise.com/xraychallenge/test_data.zip.
Download five files: pretrained_mimic_diffusion, mimic_models, paired data, images_transunet_minus1/unet-6v-latest.pt, images_transunet_padchest_minus1/unet-6v-latest.pt from https://pan.baidu.com/s/1oPoYeFsia3ngIsrSEWgX_w?pwd=refg (refg). Put them into our root dir. You may change root path name as needed.(Our root is /hhd/1/dkm/)
1. Stage1
Run the following command to perform stage1 training on ControlNet to train a rib constraint generative model
python stage1_train.py
Run the following command to perform stage1 testing on ControlNet to generate paired normal/diseased data
python stage1_generate_paired_image.py
Run the following command to perform stage1 post-precessing paired normal/diseased data
python minus.py
python gray2hotmap.py
2. Stage2
Run the following command to perform stage2 training on Transunet(trained on train split of Generated paired data) to train a pathology localization and segmentation model
python stage2_train.py
Run the following command to perform stage2 testing on Transunet(inference on test split of Generated paired data) to test a pathology localization and segmentation model
python stage2_test_generateddata.py
Run the following command to perform stage2 testing on Transunet(inference on test split of CheXpert dataset) to test a pathology localization and segmentation model
python stage2_test_chexpert.py
Run the following command to perform stage2 testing on Transunet(inference on test split of ChestX-Det-10 dataset) to test a pathology localization and segmentation model
python stage2_test_chestdet.py
Run the following command to perform stage2 testing on Transunet(inference on test split of PadChest dataset) to test a pathology localization and segmentation model
python stage2_test_padchest.py
If you have any question, please feel free to contact.