Previously I have used minimum norm imaging along with dSPM and constraint dipoles for my source localization analysis. I did group analysis and got results. Now I want to do source analysis using Maximum Entropy on the Mean (MEM) method and compare the results with the minimum norm. I have a question as you suggested in this link https://neuroimage.usc.edu/brainstorm/Tutorials/Workflows
before averaging source maps we need to normalize them individually and then rectify and smooth them. My question is do we still need to do these steps while working with cMEM source maps?
I have also another question my cMEM source maps have a pA.m unit, whereas in the BEst algorithm page all the shown maps have no units, is my maps correct?
Thank you so much
Yes, it's normal to have a unit (pA.m) when using cMEM. The figure in the tutorial should also have a unit (it's a previous bug that has been corrected since the tutorial has been made).
Thank you @edelaire , my main question however was do we need to rectify and smooth maps before averaging while working with cMEM source maps?
I will ask during our MEM meeting Monday. But I am not sure to understand the question. The question, is for group analysis, should you normalize and smooth the individual map before averaging to get the map at the group level?
If yes, it seems that the reason explaining why you would do it for MNE also applies to MEM. In both cases, you want to have a similar range of values so you normalize, and want to tackle the inter-individual difference of anatomy so smooth. But I will confirm with more experimented people
Thanks a lot! Yes exactly my question is " for the group analysis, should we normalize and smooth the individual maps before averaging them to get the map at the group level?" Also this question applies to the time that we want to do statistics between conditions and groups, " should we normalize and smooth the individual maps before doing within-subject and between-subject statistics?"
So the conclusion is that yes, you should smooth the individual maps and might want to normalize them.
The issue with normalization is that as the coefficient doesn't follow a normal distribution, the standard deviation might behave weirdly, so I think you need to try and see how the map get normalized. Unfortunately, we haven't investigate that much group analysis using MEM as we are more interested by subject analysis.
Thank you so much, I will check.