Demo code to replicate the simuations performed in A Bayesian beamformer with correlated priors: application to hippocampal source reconstruction
The code for performing the source reconstruction with the DAiSS Toolbox is all within the SPM repository.
The EBB solver in DAiSS can be interfaced with in two ways.
Like most of SPM and its respective toolboxes, DAiSS is typically interfaced with using the MATLAB Batch, an exmple of it can be seen in here but in short you can call the options using
matlabbatch{JOB}.spm.tools.beamforming.inverse.plugin.ebb.OPTION = VALUE;
It is posible to directly call the function, though its not really advised to in this way.
BF.inverse = bf_inverse_ebb(BF,S);
Where BF is the stucture DAiSS uses to store all processing and S is a structure containing all options.
DAiSS is comically undocumented, the following options for the EBB code can be called:
keeplf: (default: false) Keep oriented leadfields.
corr: (default: false) Uses correlated source priors. If no pairs matrix is suppled in the pairs option, then for every source the homolog in the opposite hemisphere is located for pairing.
pairs: (default: empty string) .mat file containing a triangular matrix (full or sparse) with non-zero elements representing which sources are correlate with each other. For example if sources 3 and 5 are correlated then the matrix element (3,5) or (5,3) should equal 1. To keep a source uncorrelated set its respective diagonal element to 1. (e.g. source 4 is uncorrelated so (4,4) = 1).
iid: (default: false) Set the prior matrix to an identity matrix, which is the equivalent of a basic minimum-norm inversion.
noise: (default: empty string) BF.mat file of empty room noise (if an i.i.d noise assumption is not to your liking).
reml: ('strict' | 'loose' [default]) options for the ReML optimising algorithm, setting to strict will try and balance the weighting of the noise to source matrices with a 1:20 ratio (similar to a 5% regularisation). Loose allows it to freely optimise in the same way as SPM.
If you find the correlated prior EBB algorithm useful. Please cite the following.
O’Neill, G.C., Barry, D.N., Tierney, T.M. et al. Testing covariance models for MEG source reconstruction of hippocampal activity. Sci Rep 11, 17615 (2021). https://www.nature.com/articles/s41598-021-96933-0