Email discussion lists for the UK Education and Research communities

## SPM@JISCMAIL.AC.UK

#### View:

 Message: [ First | Previous | Next | Last ] By Topic: [ First | Previous | Next | Last ] By Author: [ First | Previous | Next | Last ] Font: Proportional Font
 LISTSERV Archives SPM Home SPM May 2011

#### Options

Subject:

spatial FWHM values for AFNI's 3dClustSim

From:

Date:

Mon, 9 May 2011 14:26:10 -0400

Content-Type:

text/plain

Parts/Attachments:

 text/plain (55 lines)
 ```Dear SPM/fMRI experts, On my SPM8 T-maps I am performing cluster-size thresholding using 3dClustSim of the AFNI package (formerly known as AlphaSim), which requires an estimation of the correlation of the voxels in a 3D dataset. I use a tool of the same package (3dFWHMx) to estimate the degree of voxel spatial correlation, based on an algorithm by Forman and colleagues (http://onlinelibrary.wiley.com/doi/10.1002/mrm.1910330508/pdf) However, a colleague uses a different approach by extracting the FWHM values for the 3 axes from the SPM.mat file - SPM.xVol.FWHM. I tried both methods and noticed that they produce different results - here is an example for one contrast where the values don't seem to differ that much, but sometimes the differences are greater than 1.0: *** 3dFWHMx: FWHMx = 9.82782 FWHMy = 10.0744 FWHMz = 9.31861 *** SPM.xVol.FWHM: FWHMx = 10.2832 FWHMy = 10.3741 FWHMz = 9.4135 My questions are: (1) Do the values in the SPM.mat file even reflect the voxel spatial correlation of the contrast/the T-map? (2) If so, what algorithm is their estimation based on (or: are they based on a different calculation than Forman et al.)? and (3) What would be the recommended method? Last but not least: (4) Is there an (SPM-based) alternative method for cluster-size thresholding? Thanks in advance! All the best, Franziska -- Franziska M. Korb, Ph.D. Postdoctoral Associate (Egner Lab) Center for Cognitive Neuroscience LSRC Building, Room C03D, Research Dr Box 90999 Duke University Durham, NC 27708 phone: ++1-919-684.1034 http://sites.google.com/site/egnerlab/ ```