Modified connectome-based predictive model. See preprint of book chapter outlining this method: https://osf.io/ay95f/
- Enables inclusion of covariates at feature selection or model-building stage or both stages
- Provides ability to threshold edges based on p-values (e.g. only select edges with p-values for correlation between edge and target variable < .01) or sparsity (e.g. select top 7.5% of edges based on correlation bewteen edge and target variable)
- Permits different choices of k-fold cross-validation schemes
- Returns weights (i.e. regression coefficients) from trained model for external validation on other datasets
- Creates masks for visualisation in bioimagesuite and enables creation of masks when restricted connectivity matrices are used (i.e. when nodes are removed from the original 268 x 268 connectivity matrix)
- Enables leave-site-out cross-validation
- Optimises CPM for larger datasets via the use of parallel computing
Functions with prefix "CPM__" are adapted/generalised functions of code written by Xilin Shen and Emily Finn and is uploaded here with their permission (codeshare_behavioral_prediction.m as obtained here: https://www.nitrc.org/frs/download.php/8071/nn_code.zip). This code is therefore copyrighted to Xilin Shen and Emily Finn as detailed below:
Copyright 2015 Xilin Shen and Emily Finn This code is released under the terms of the GNU GPL v2. This code is not FDA approved for clinical use; it is provided freely for research purposes. If using this in a publication please reference this properly as: Finn ES, Shen X, Scheinost D, Rosenberg MD, Huang, Chun MM,Papademetris X & Constable RT. (2015). Functional connectome fingerprinting: Identifying individuals using patterns of brain connectivity. Nature Neuroscience 18, 1664-1671.
All other functions and scripts (e.g. functions with suffix "_CPM" - including run_flexible_CPM_leaveSiteOut) are original and this code is also released under the terms of the GNU GPL v2
Code is written and tested on MATLAB R2020a