Individual difference study AND GRF-correction

Submitted by l5583325 on Sun, 09/14/2014 - 00:59

Hi, everyone!

I am doing a whole-brain seed-based functional connectivity analysis using pre-defined ROI. My aim is to find connectivity which is correlated with a individual difference data.

All of the proprocess and statistic steps were performd by Dpabi.

My results can survivive in GRF-correction (voxel p<0.025, cluster p<0.1, two-tailed, I got this from video course). However, previous papers usually use this method for group difference dectection (Z>2.3, cluster p<0.05).

Therefore, I have two questions here:

(1) Does it right to use GRF in correlation analysis?

(2) Does their use one-tailed way in setting (Z>2.3, cluster p<0.05). So voxel p<0.025,cluster p<0.1(two tailed)=Z>2.3, cluster p<0.05(one tailed)?

(3) The GRF correction (voxel p<0.025,cluster p<0.1,two tailed) in my VBM result need 640 voxels, however, in my GCA result, it needs only 74 voxels. Some thing is wrong here?

Forums:

YAN Chao-Gan

Tue, 09/16/2014 - 18:48

1. Yes.

2. Yes.

So voxel p<0.0214, cluster p<0.1 (two tailed)  =  TWICE: Z>2.3, cluster p<0.05 (one tailed).

3. Because your VBM results are on voxels size of 1*1*1? And the smoothness is even big?

Best,

Chao-Gan

 

l5583325

Wed, 09/17/2014 - 01:24

Thanks for reply!

My VBM is also analysised by default setting in Dpabi(smoothness=8) and  used whole brain gray matter group mask created by averging all the subjects gray matter volume  when the software asked for a mask.

My GCA resutls is generated by GCA function in REST (3*3*3, I think) and used 90% mask created by dpabi.

Then I tried several resting state analysis in GRF, the standard never above 200. I want to know whether the number( 642 voxel )in VBM has some problems? 

Best Wish

Wei Liu

YAN Chao-Gan

Wed, 09/17/2014 - 15:03

1. 642 * 1*1*1 = 642 mm3.

    74*3*3*3 = 1998 mm3.

Which cluster threshold is bigger?

2. VBM is smoothed by 8mm, your functional images are smoothed by 4mm? This makes a difference as well.

Best,

Chao-Gan