Network Constraints on Longitudinal Grey Matter Changes in Psychosis

Reference

Chopra, S., Oldham, S., Segal, A., Holmes, A., Sabaroedin, K., Orchard, E. R., … & Fornito, A. (2022). Network constraints on longitudinal grey matter changes in first episode psychosis. medRxiv.

---

Background

Different regions of the brain’s grey matter are connected by a complex structural network of white matter fibres, which are responsible for the propagation of action potentials and the transport of trophic and other molecules. In neurodegenerative disease, these connections constrain the way in which grey matter volume loss progresses. Here, we investigated whether connectome architecture also shapes the spatial pattern of longitudinal grey matter volume changes attributable to illness and antipsychotic medication in psychosis.

---

Code and Data release

The scripts folder contains the three primary analysis and visualisation scripts:

• runNDM.R begins with a function which takes in a vector of volume change values (or any statistic) for each region, a structural connectivity (SC) matrix, optionally a functional connectivity (FC) matrix and applies the Network Deformation Model (NDM). The function allows the model to optionally be unweighted or weighted by SC or FC. The function also computes p-values based on the three different null models (connectome, spin and parameterised nulls). Note: Due to data space restrictions the parameterised nulls are not included in this repository.

• plotNDM.R script takes the output of runNDM.R function (a .RDS file) and plots the results like below:

• runNDMepicenter.R runs the epicenter analysis using the same inputs as runNDM.R and plots the results.

The functions folder contains a collection of helper functions for the primary analysis and visualization scripts. The majority of the functions are used to generate group-level FC/SC matrices and do not need to be rerun as the outputs are provided in the contents data folder described below.

The data folder contains:

• The representative group-consensus SC and group-mean FC matrices for healthy control data sets. The individual-level SC data was processed using tractoflow and filtered using COMMIT2 prior to the group-consensus matrix being computed (please refer to paper for details). The FC data was processed using fmriprep and denoised using AROMA, with and without Global Signal Regression (please refer to paper for details), prior to a group average matrix being computed.

• The Deformation Based Morphometry (DBM) voxel-level contrast t-statistic maps for all contrasts. HC = Healthy control, PIPT = Placebo group and MIPT = Medication group. (please refer to paper for details on processing).

• The atlases used (Schaefer 300 parcel, 7 network and Tian/Melbourne atlas subcortical scale II) for volume, SC and FC estimates.

---

Questions

Please contact me (Sidhant Chopra) as [email protected] and/or [email protected]