Particle filters and RAP-MUSIC in MEG source modelling: A comparison

We compared the novel particle filtering approach to an established MEG inverse modelling algorithm “recursively applied and projected (RAP) MUSIC”. Both methods are able to reconstruct temporally correlated sources in an automatic manner. We evaluated the performance of the two methods in critical yet neurophysiologically plausible source configurations. RAP-MUSIC outperformed particle filters when modelling sources of low signal-to-noise ratio or short duration, while the particle filtering approach performs better when reconstructing correlated sources or sources with positions and orientations varying across the response.