Hi Abigail,
Regarding the different options for selecting the mask:
1. During spatial preprocessing - we used spm defaults
2. during fMRI model specification - this is our preferred method, since you have the activation already saved in the con images.
3. during results inquiry - I have not personally tried this method but I think if you were to go with it you may have to save out a new activation map for each subject if you want to conduct 2nd level analyses. The mask.img that you are talking about is generated by SPM during model estimation and displays the search region of your first level analysis. So it would not make much sense to use this mask to generate the mask.
So going back to your first question, I definitely think method #2 is the way to go. First make a decision about what you want your defaults mask threshold to be (-inf to .8). Then, binarise SPM's brainmask.nii (found in spm/apriori directory) using SPM's ImCalc function. Your formula input would be "i1> <your threshold here>." The ImCalc function will basically give you an output image in which all values lower than your threshold are set to zero and all values above the threshold are set to one. When you specify this mask in model specification, the estimation will only look at voxels designated as ones. As Minh suggested, you don't want skull included in your search region, so do a check reg of your binarised image with a group structural image and see if the region is inclusive of most brain regions but excludes skull.
-Vy
________________________________________
From: Abigail Livny-Ezer [[log in to unmask]]
Sent: Tuesday, September 28, 2010 7:53 AM
To: [log in to unmask]; Vy Dinh
Cc: Abigail Livny-Ezer
Subject: Re: brain mask
Hi Vy
Thanks for your reply!
I think the first option you mentioned, selecting SPM's brainmask is the best option for me.
However, I am not sure which IMG I need to choose and in which module?
As I wrote before there are a few options in SPM8 to add a mask.
One, in the spatial preprocessing in the segmentation, second in the fMRI model specification module, in which it is possible to add an explicit mask and third in the results module (mask with image)
I still did not understand from your answer which one is the correct place to add a mask which will exclude voxels outside the brain. And which image should I use as the mask?
SPM creates a file called mask.img but it is created only later in the model estimation so I can't call it earlier.
Which image did you mean when you wrote: " SPM's brainmask"?
Also, you mentioned that if I wish to use the -inf threshold, I should convert this to a binarised image since the brain mask is smoothed, how do I do this?
Thanks for your help,
Abigail
|