I'd like to use AlphaSim to correct VBM p values. The voxel size is 1.5*1.5*1.5, and smoothing used a Gaussiankernel of 8mm full width at half maximum. I wondered what the rmm and cluster size shoud be set.

rmm = 2 [i.e., <1.5*sqrt(2)] means face connection.

rmm = 2.5 [i.e., <1.5*sqrt(3)] means edge connection. This is SPM's default.

rmm = 3 (i.e., >1.5*sqrt(3)] means corner connection. This is FSL's default.

For smooth kernel, the effective smooth kernel may not be the same as your applied smooth kernel (8 8 8). You'd better use AFNI's 3dFWHMx or FSL's smoothest to estimate the effective smooth kernel on the residual file. -- This strategy will be automated in the next release of our software.

Just to follow up on this question, I wonder whether we could use rest_Smoothest.m in REST_V1.8_130303 to directly estimate the smooth kernel (e.g., using t map as input)?

I wrote that program to be equivalent to FSL's smoothest, and the entire (GRF correction) procedure (input T maps) to be equivalent to FSL's easythresh.

However, this procedure (easythresh as well) usually overestimates the smooth kernel than estimating the smooth kernel on the residuals. Thus, please keep in mind this procedure is pretty strict.

YAN Chao-Gan

Sat, 10/26/2013 - 13:00

Permalink

## Re: [RFMRI] AlphaSim

zhongxu

Sun, 10/27/2013 - 01:50

Permalink

## Re: [RFMRI] AlphaSim

YAN Chao-Gan

Sun, 10/27/2013 - 02:50

Permalink

## Re: [RFMRI] AlphaSim