CausalGCN-PDPA

This repository holds the code for the paper

A Causal-Driven Graph Convolutional Networks for Postural Abnormality Identification in Parkinsonians

All the materials released in this library can ONLY be used for RESEARCH purposes and not for commercial use.

The authors' institution (Biomedical Image and Health Informatics Lab, School of Biomedical Engineering, Shanghai Jiao Tong University) preserve the copyright and all legal rights of these codes.

Author List

Xinlu Tang, Rui Guo, Chencheng Zhang, Xiahai Zhuang*, and Xiaohua Qian*

Required

Our code is based on Python3.9 There are a few dependencies to run the code. The major libraries we depend are

• PyTorch1.10.0 (http://pytorch.org/)
• numpy
• pandas

Set up

conda install --yes --file requirements.txt

Attention: Please run this project on linux. In different pytorch environment, the model may obtain different results.

Quickly start

run the main.py by this command:

python main.py --config ./train_causal_pre.yaml

The results will be saved in ./result folder.

Description of input data structure

Data loading: FeederGraph in /tools/feeder.py

Shape of input data: [B,P,d]. B: batch size, P: number of patches within a sample or number of nodes within a graph, d: length of node feature vector.

Input data acts as the node feature matrix, and patch coordinate information is needed additionally.

Adjacency matrix calculation: ConstructGraph in /tools/construct_ graph_simple.py

Shape of output matrix:[B,P,P], formed by stacking the adjacency matrices of all graphs along the first dimension.

For the data from users, the adjacency matrix stack can be directly loaded in or be calculated through a customized function. Finally, the node feature matrix and the adjacency matrix stack are fed into the model.

Comparing methods

CNN+MLP

cd ./ComparingMethods/CNN_mlp
python main_cnnmlp.py --config ./train_cnnmlp.yaml

Radiomics

Reference: https://pyradiomics.readthedocs.io/en/latest/

cd ./ComparingMethods/radiomics
python radiomics_extract.py
python main_radiomics.py --config ./train_radiomics.yaml

DAMIDL

Reference: https://github.com/WyZhuNUAA/DA-MIDL/

cd ./ComparingMethods/DAMIDL
python main_DA.py --config ./train_DA.yaml

BrainGNN

Reference: https://github.com/xxlya/BrainGNN_Pytorch

cd ./ComparingMethods/BrainGNN
python main_braingnn.py --config ./train_braingnn.yaml

GCN

cd ./ComparingMethods/gcn
python main_gcn.py --config ./train_gcn.yaml

Contact

For any question, feel free to contact

Xinlu Tang : tangxl20@sjtu.edu.cn