topological-profiles

Code for building topological profiles of neurological diseases from brain regional information (from MRI or PET imaging)

05/2023: major update of the code to run with the released version of GPPM (https://gitlab.inria.fr/epione/GP_progression_model_V2)

Note

• The topological profile code is now implemented in python (3.7) for a smoother interface with GPPM: https://gitlab.inria.fr/epione/GP_progression_model_V2.
• It has a number of standard requirementes (numpy, pandas, torch, matplotlib) and bctpy: https://pypi.org/project/bctpy/.
• Also, it requires GPPM to be locally in your computer, as it uses some of its data loading procedures for loading GPPM data.

Usage

• Code can be run from the main.py script; please note that it will require (standard) saved output from GPPM in the form of a "model_.pth" and "tr_.pth" files.
• By default it runs with some (toy) example of ADNI data, for which GPPM has produced a set of output which can be found in the "GPPM_data" folder
• By default it computes a set of 9 graph theory metrics on the structural connectome/cortical distance;
  • weighted degree
  • betweeness centrality
  • closeness centrality
  • clustering coefficient
  • inverse of weighted degree
  • inverse of clustering coefficient
  • shortest path to epicenter (via connections)
  • shortest path to epicenter (via cortical paths - also called "cortical proximity")
  • constant, homogeneous, isotropic propagation
• The folder "connectome_data" contains some pre-computed average structural connectome and average cortical distance data on a set of HCP brains: https://www.humanconnectomeproject.org/ (for how these were computed, see Oxtoby et al, Frontiers 2017 and Garbarino et al, eLife 2019)

---

Older version (Matlab)

The previous version of the code, running on Matlab, can be still found in the "matlab" branch. Please note that it will run smoothly on dummy data, but IT WILL NOT EASILY RUN if you have data obtained with the released version GPPM: https://gitlab.inria.fr/epione/GP_progression_model_V2. Main.m launches the code on dummy ADNI data.

It requires:

• "Brain Connectivity Toolbox" (https://sites.google.com/site/bctnet/);
• "npy-matlab" (https://github.com/kwikteam/npy-matlab)