L1-norm and L2-norm neuroimaging methods in reconstructing extended cortical sources from EEG

We have previously reported a new sparse neuroimaging method (i.e. VB-SCCD) using the L1-norm optimization technology to solve EEG inverse problems. The new method distinguishes itself from other reported L1-norm methods since it explores the sparseness in a transform domain rather than in the original source domain. In the present study, we conducted a Monte Carlo simulation study to compare the performance of VB-SCCD and other two popular L2-norm neuroimaging methods (i.e. wMNE and cLORETA) in reconstructing extended cortical neural electrical activations. Our simulation data suggests that the VB-SCCD method is able to reconstruct extended cortical sources with the overall high accuracy. It has significantly higher accuracy, less number of false alarms and less number of missing sources when studying complex brain activations (up to 5 simultaneous sources). This new sparse neuroimaging method is thus promising to have many valuable applications in neuroscience and neurology problems. It is also applicable to MEG neuroimaging.