Data-Driven Parkinson’s Disease Subtyping and Digital Biomarker Discovery by Correlating Video-Captured Gait with rs-fMRI Brain Connectivity
This repository contains code for a data-driven Parkinson's disease subtyping approach using motion and brain functional connectivity.
The code has been tested on Linux operating systems (Ubuntu 20.04). All package requirements are located in requirements.txt.
Run the following commands with Python 3.7:
git clone https://github.com/markendo/PD-Subtyping-Motion-Brain.git
cd PD-Subtyping-Motion-Brain
pip install -r requirements.txt
The package should take a few minutes to install on a standard computer.
We have provided a notebook containing subtype analysis on a toy dataset. The expected output is included in the notebook.
We also include notebooks containing the workflows for generating full subtype analysis results and fMRI subnetwork analysis results.
To run on your own data, first extract motion encoder outputs from this repo and place them at data/motion_encoder_outputs. Corresponding FC data should be stored at data/fc_data.