Kqp1227 / GSKET_FFA

Knowledge Matters: Radiology Report Generation with General and Specific Knowledge

结合通用知识和特定知识的医学报告生成

Requirements

• Python >= 3.6
• Pytorch >= 1.7
• torchvison
• Microsoft COCO Caption Evaluation Tools
• CheXpert

conda activate tencent

Data

Download IU and MIMIC-CXR datasets, and place them in data folder.

• IU dataset from here
• MIMIC-CXR dataset from here

Folder Structure

• config : setup training arguments and data path
• data : store IU and MIMIC dataset
• misc : build general knowledge and specific knowledge
• models: basic model and all our models
• modules:
  • the layer define of our model
  • dataloader
  • loss function
  • metrics
  • tokenizer
  • some utils
• preprocess: data preprocess
• pycocoevalcap: Microsoft COCO Caption Evaluation Tools

Training & Testing

The source code for training can be found here：

main_basic.py: basic model

main.py: model with knowledge

The values of all the hyperparameters can be found in the folder 'config'.

To run the command, you only need to specify the config file and the GPU ID and iteration version of the model to be used

Example： python main.py --cfg config/iu_retrieval.yml --gpu 0 --version 1

Citation

Shuxin Yang, Xian Wu, Shen Ge, S. Kevin Zhou, Li Xiao, Knowledge Matters: Radiology Report Generation with General and Specific Knowledge. Medical Image Analysis，2022

Contact

If you have any problem with the code, please contact Li Xiao([e-mail]) or Shuxin Yang([e-mail]).