What kind of limitations are you referring to? You can create a mask covering 5 regions in the frontal lobe, or 5 separate masks. Both are OK. If you are doing ROI analysis with VBM8 or CAT12 built for SPM, remember to reduce the stat significant threshold to p = 0.05/5, i.e. p = 0.01, if you test 5 regions separately.

Dear Andy,

Thank you for your response. I am exploring the plasticity of the brain of a group for the first time using CAT 12. Based on previous vbm analysis from literatures, I got 32 regions that I need to create masks. If I divide regions according to the lobes it will be around 5-6 regions in each lobe and subcortical structures separately. I am planning to create masks sperately for each regions. In such case how much the p value should I give? The p value that you mentioned is the corrected or uncorrected?

Thanks

Ramesh

It's the principle of Bonferroni correction. So if you have 32 test regions, meaning that you run the same test 32 times, the significance level should theoretically be p = 0.05/32. And I guess you should look at the FDR or FWE p, if you are running voxel- or cluster- level inference. If you are extracting the volume of the whole ROIs using Marsbar or something, and run tests in SPSS, then it's the p in the stat program.

Dear Andy,

Thank you very much for your response. Recently I read a response SPM post ( https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1906&L=SPM&P=R62627&X=A8A29F6466DBB43C37&Y=mgrameshbabu2013%40gmail.com ) in which it is mentioned that p = 0.001/ number of mask. Here I am getting confusion. Which is the correct one?