TriNet-DCM

A spectral DCM pipeline for resting-state fMRI effective-connectivity analysis and dementia conversion prediction using OASIS-3 data. This work builds on the UKB_DCM_dementia codebase:

https://github.com/Wolfson-PNU-QMUL/UKB_DCM_dementia/

Citation

If you use or extend this repository, please cite:

Ereira, S., Waters, S., Razi, A. et al.
Early detection of dementia with default-mode network effective connectivity.
Nature Mental Health 2, 787-800 (2024).
https://doi.org/10.1038/s44220-024-00259-5

What's included

All scripts are in Code/:

• trinet_pipeline.m - main entry point; run one stage at a time
• trinet_structural_MRI_preprocess.m - T1 preprocessing (SPM12 segmentation + normalization)
• trinet_functional_MRI_preprocess.m - rs-fMRI preprocessing (realignment, slice-timing, coregistration, normalization, smoothing)
• trinet_extract_timeseries.m - ROI time-series extraction
• trinet_firstlevelDCM_QC.m - first-level spectral DCM fitting + QC
• trinet_EC_classifier.m - second-level PEB/BMA + elastic-net classifier
• trinet_dem_EC_prognosticator.m - second-level PEB/BMA + elastic-net prognosticator
• Helper scripts: trinet_ROI_specify.m, trinet_BuildRegressors.m, trinet_general_classifier.m, trinet_general_prognosticator.m

Requirements

• MATLAB R2021b or later
• SPM12 (tested on v7771)
• glmnet for MATLAB

Data requirements

This repo does not include data. The scripts expect:

• rs-fMRI and T1 images arranged with fMRI/ and T1/ inside each subject folder
• CSV files containing metadata (e.g. EID, TR, TE, DEM_STATUS, MCI_STATUS, age_at_scan, EDUC, etc.)

Quick start

Open Code/trinet_config.m and edit the paths:

• SPM + glmnet locations
• fMRI/T1 directory
• Metadata CSV paths
• Output directory

Launch the pipeline in MATLAB:

trinet_pipeline

When prompted, choose which stage (1-6) to run.