Signal Reconstruction

The signal reconstruction stage produces the estimates of the activation of ROI voxels (source dipoles) in time. ROIs need to be defined by the name and coordinates expressed in the standard MNI space. In the case of individual head and source models, these standard coordinates are individually warped before the signal is reconstructed. See the following code:

loc.ROI(1).name = 'dACC';   
loc.ROI(1).pos = [7 18 33];  

loc.ROI(2).name = 'raIns'; 
loc.ROI(2).pos = [37 16 -2];  

loc.ROI(3).name = 'rDL';  
loc.ROI(3).pos = [58 18 44];

loc.ROI(4).name = 'rVL';  
loc.ROI(4).pos = [ 42 25 14];  

loc.ROI(5).name = 'rPPC';  
loc.ROI(5).pos = [48 -52 50];  
 

The results of reconstruction are saved as two files: 7_SOURCEACT/DCx/unc/…sigroi_unc.mat and 7_SOURCEACT/DCx/lc/…sigroi_lc.mat. The former contains a direct superposition of all brain ICs in selected cortical voxels (ROI dipoles) and it is saved for reference only. The former file contains ROI activations after the leakage correction procedure.

Leakage correction

The source reconstruction process tends to produce sources that are spatially blurred when compared to the original source extent. This inaccuracy is imminent due to the fact that we reconstruct a high number of cortical voxels using only a limited number of measurement channels. This may cause the reconstructions of two closely located sources to overlap, which will introduce some zero-lag correlation.

To remove this artificial instantaneous relationship between signals, a symmetric multivariate orthogonalization procedure is performed by default using the ROInets toolbox. This step is important; only the leakage-corrected signal should be used for further analysis, as the uncorrected signals are contaminated with spurious connectivity.

During the signal reconstruction, the con.decimate parameter enables downsampling the reconstructed signal by providing the factor for sample frequency division. It is intended to set the resulting sampling rate near the recommended value of 128 Hz (for example, if the sampling rate after preprocessing is 256 Hz, this parameter should be set to 2 at this point).