Dear Zoya,
I know that this might be confusing. However, this is caused by the correction for nonstationarity that considers different local smoothness. Because Random Field theory largely depends on smoothness of your data (i.e. based on the residuals) different local smoothness can lead to under- or overestimation of cluster size and correction for nonstationary clsuters exactly corrects for this effect. Therefore, smaller clusters can survive the size threshold if local smoothness is smaller than average. See Hayasaka & Nichols for more details: https://doi.org/10.1016/j.neuroimage.2003.08.003
https://doi.org/10.1016/j.neuroimage.2004.01.041
Best,
Christian
On Tue, 30 Nov 2021 10:24:05 +0000, Zoya Mooraj <[log in to unmask]> wrote:
>Hello,
>
>I have a question regarding the 'Transform SPM Maps' option in the CAT12 toolbox.
>
>As a bit of background: I am looking at longitudinal data from two groups. I preprocessed them through the CAT12 longitudinal pipeline and used a flexible factorial model to compare my two groups (G1=control and G2=training). There are four timepoints (T1 - T4) each, so my design matrix looks like:
>G1T1 G1T2 G1T3 G1T4 G2T1 G2T2 G2T3 G2T4. I then created the following contrast [0 0 0 0 -3 -1 1 3] to test for an increase in volume in the training group.
>
>I then used the 'Transform F-Volume Maps' cat12 option, converting the F-value to p, setting the peak-level threshold type to uncorrected, with a .0001 threshold, and with a cluster extent threshold set to k=100.
>
>However, once I look at the output that gets printed to the matlab console with the associated regions (attached), amongst the remaining clusters there is one with a size of 73 voxels. This was surprising to see since the threshold was set to 100, so I would expect only clusters 100 voxels or more to be included. Is there any reason why this would happen, or am I perhaps missing something in this process?
>
>Thank you very much!
>
>Best,
>Zoya Mooraj
>
>P_Linear_Increase_p0.01_k100.nii: Positive effects
>
>
>Neuromorphometrics
>______________________________________________________
>
>F-Value Size xyz [mm] Overlap of atlas region
>
> 54.56 198 24 48 2 98% Right Cerebral White Matter
> 2% Right MFG middle frontal gyrus
> 46.54 260 17 24 -5 59% Right Cerebral White Matter
> 31% Right Caudate
> 7% Right Lateral Ventricle
> 3% Right Putamen
> 33.44 104 -36 -36 24 99% Left Cerebral White Matter
> 32.61 100 -33 35 2 85% Left Cerebral White Matter
> 9% Left OrIFG orbital part of the inferior frontal gyrus
> 6% Left TrIFG triangular part of the inferior frontal gyrus
> 31.77 122 -15 24 -5 42% Left Cerebral White Matter
> 31% Left Lateral Ventricle
> 27% Left Caudate
> 25.42 73 14 38 8 99% Right Cerebral White Matter
> 1% Right ACgG anterior cingulate gyrus
> 24.50 535 -11 8 0 38% Left Cerebral White Matter
> 27% Left Pallidum
> 24% Left Putamen
> 10% Left Caudate
>
>
>
|