How to determine REST AlphaSim parameters properly

Dear Experts,

I have read a lot of the supporting documentation related to REST AlphaSim, along with many posts in the online forum; and although I feel I have a good understanding of the procedure I have two outstanding questions which my reading has not answered. I'd really appreciate your help in these matters.

1) How to determine the 'rmm' parameter to input properly.
I understand that this is the acceptable distance between the centre of voxels that can be counted as belonging to the same cluster. Therefore I have assumed that this relates to the smoothing kernel I applied [8 8 8], and have inputted 8. Is this correct?

2) Using the GUI to Estimate the FWHM from inputting a statistical map.
I have been enterring an SPMt file, which I read on the online forum is ok to do so. However, depending on which SPMt I used (according to the specific t contrast) I get different values as the estimate output. If this estimate influences the simulation output overall then the results of the simulation would only be valid for that particular ttest, and I would need to repeat this process every time I used a new contrast. This seems illogical to me, perhaps someone could explain it please?

Thanks in advance for the help,
Sincerely,
Gemma.

Hi Gemma,

1. rmm is used to define connection cretiria. For example, when voxel size = 3*3*3, rmm = 4: face connection; rmm = 5: edge connection (SPM use this); rmm = 6 corner connection (FSL use this).

2. Use the statistical image to estimate the smoothness is a roughly approach. A better way is using the residual, thus all the contrast should use the same smoothness. SPM will estimate the smoothness by itself, and stored in SPM.mat. Probably in SPM.xVol.FWHM?

Best,

Chao-Gan

Dear Chao-Gan,

Many thanks for your very helpful response. With your information I have been able to find the smoothness estimation created by SPM and will use this in my AlphaSim computation. You also directed me to find out more about the connectivity criteria which was very helpful, however despite my further reading on this I was not able to apply this and the information you gave me to my own example - perhaps you can help further?

I am using AlphaSim with a 2 dimensional data set, and my dimensions are: 2.13mm x 2.69mm (x1). I computed the length along the diagonal of this pixel, and intend to use this as the rmm. Is it correct that this value could be used as a possible rmm (relating to the corner connection I guess)?

Furthermore, it may be preferable to use edge connection instead, but I could not understand how you got from voxel size 3x3x3 to rmm=5 in your example? Perhaps you can explain?

Thanks again and Best Wishes,
Gemma.

Hi Chao-Gan,

I have a few questions regarding your post.
1. You mentioned that SPM uses edge connection whereas FSL uses corner connection. Which connection criterion is more commonly used and why?
2. You also mentioned that it is better to use the residual (I believe, it's the ResMS.img) than the SPMt file, so that all the contrasts would have the same smoothness. Could you elaborate on that? In my case, I have two statistical images from a VBM analysis, one for Year 1 and the other for Year 2. Should I use the residual or t images to estimate the smoothness?

Thanks!

Best,
Darren

 Dear ZHANG_RESTadmin,

Many thanks for your reply and thanks for our suggestion.

I have only used limited functions of the software and therefore I do not have a file called rest_AlphaSim.m. Perhaps I can explain a litle more about what I've done...

I am only using the REST AlphaSim function within the Statistical Analysis function. I am using this to estimate the cluster size required to indicate genuine scalp activity in my EEG data. I enter the mask used in the statistical analysis of my EEG data (performed later in SPM) which is a two dimensional image with 574340 pixels, with each pixel measuring 2.13mm x 2.69mm. 

I would like some help to know what to enter into the rmm field to create an accurate measure of the cluster size required to identify genuine scalp activation using the edge detection rmm criteria. 

I would really appreciate any help with this, as I require this to finalise some results that are othewise ready for submission.

Many thanks,
Best Wishes,
Gemma.