Topological and geometric signatures of brain network dynamics in Alzheimer’s disease (Alzheimer's & Dementia Journal)

This repository contains the code accompanying the paper:

Topological and geometric signatures of brain network dynamics in Alzheimer’s disease

🔗 Read the full open‐access paper here: https://doi.org/10.1002/alz.70545

Overview

This work presents a novel permutation-based framework to identify topological and geometric biomarkers of Alzheimer’s disease (AD) using dynamic functional connectivity (dFC) derived from resting-state fMRI data.

The repository includes:

• ✅ Data Preprocessing
• ✅ Dynamic Connectivity Matrix Construction
• ✅ Distance-Based Time Series Generation
• ✅ Permutation Testing Across Diagnostic Groups and Sexes

Folder Structure

├── Data-Preprocessing/
│   ├── fmriprep_simg.sh             # Runs fMRIPrep using Singularity
│   ├── generate_conn_matrices.py    # Generates connectivity matrices from preprocessed data
│   └── process_raw.sh               # Processes raw BIDS-formatted data
│
├── ConnectivityPipeline/
│   ├── extract_distance.py          # Computes distance metrics (e.g., Wasserstein, spectral)
│   └── Process_sliding_window.py    # Implements sliding-window dFC computation
│
├── PermutationTest/
│   ├── oasis_5peak_permutation.py   # Permutation test using 5-peak-based features
│   └── oasis_mean_permutation.py    # Permutation test using mean-based features
│
├── requirements.txt                 # Python dependencies

Requirements

This project relies on a combination of Python packages and neuroimaging tools for preprocessing, connectivity analysis, and statistical testing. SLURM job scripts are included for HPC environments.

Python Packages

To install all core dependencies:

pip install -r requirements.txt

Or install them individually:

• numpy
• pandas
• matplotlib
• seaborn
• scipy
• scikit-learn
• tqdm
• nibabel
• nilearn
• networkx

Note: Some scripts also use google.colab for mounting Google Drive.

Optional / Environment-Specific Tools

• fMRIPrep (required for anatomical preprocessing)
• Install via Singularity or Docker
• SLURM workload manager (for HPC job scheduling)
• Shell (bash) and SBATCH scripts for parallel job submission

System Requirements

• Python 3.8+
• 16–32 GB RAM recommended for full pipeline execution
• Access to BIDS-formatted resting-state fMRI data

Citation

If you use this code, please cite:

@article{yi2025topological,
  title={Topological and geometric signatures of brain network dynamics in {A}lzheimer's disease},
  author={Yi, Luopeiwen and Lutz, Michael William and Wu, Yutong and Li, Yang and Songdechakraiwut, Tananun},
  journal={Alzheimer's \& Dementia},
  volume={21},
  number={8},
  pages={e70545},
  doi = {https://doi.org/10.1002/alz.70545},
  url = {https://alz-journals.onlinelibrary.wiley.com/doi/abs/10.1002/alz.70545},
  eprint = {https://alz-journals.onlinelibrary.wiley.com/doi/pdf/10.1002/alz.70545},
  year={2025}
}

License

This project is licensed under the MIT License. See the LICENSE file for details.