In recent years, representing the brain as a network and analyzing it through machine learning and graph analytics has emerged as a powerful approach for studying its structure and function. Unlike traditional examination methods, this graph-based representation preserves both structural and positional information, offering unique insights into how different brain regions interact and influence each other.
Such methods have proven particularly valuable in investigating neurodegenerative and neurodevelopmental disorders like Parkinson’s disease and Autism. These conditions often lead to measurable alterations in brain network topology, which can be quantified through specific graph metrics — for example, closeness centrality, node degree, and clustering coefficient.
This project aims to evaluate whether graph metrics learned in the Learning from Networks course can help distinguish brain networks of individuals with Parkinson’s disease or Autism from those of healthy controls.
We use data from the publicly available dataset A Collection of Brain Network Datasets [2], focusing specifically on:
- Autism Brain Imaging Data Exchange (ABIDE)
- Parkinson’s Progression Markers Initiative (PPMI)
Brain networks are constructed from correlation matrices generated using the Automated Anatomical Labeling (AAL) parcellation algorithm [3], which divides the brain into 116 distinct regions. Pearson’s correlation coefficient [4] is used to measure the connectivity between these regions, resulting in a 116×116 adjacency matrix (13,456 undirected edges).
These datasets cover thousands of subjects with diverse age and sex distributions, enabling a broad exploration of demographic effects on brain network structures.
- Clone the repository:
git clone https://github.com/mattreturn1/Graph_Based-Brain-Networks- Download the library required
pip install -r requirements.txt- The repository contains the already filtered folders.
If you want to test thefolders_organizer.py, you need to downloadabide.zipandppmi_v2.zipfrom the following link: https://auckland.figshare.com/articles/dataset/NeurIPS_2022_Datasets/21397377
Then extract the contents of the abide.zip folder and rename the extracted folder to abide.
Similarly, extract the contents of the ppmi_v2.zip folder and rename the extracted folder to ppmi.
Move both renamed folders to the root folder of the project.
This project uses the dataset:
Xu, Jiaxing; Yang, Yunhan; Huang, David; Gururajapathy, Sophi; Ke, Yiping; Qiao, Miao; et al. (2023).
A Collection of Brain Network Datasets. The University of Auckland. Dataset.
https://doi.org/10.17608/k6.auckland.21397377.v7
Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).