Dear FSL VBM Gurus,
I have a question about making a template_list file for a VBM analysis of 241 HIV patients that had T1 structural images collected using 7 magnets at different locations in the United States. 66 of the HIV subjects have chronic neuropathic pain in their feet while 175 of the HIV subjects do not have chronic neuropathic pain.
If I were making a template_list file in the setting of all the subjects having had their scans on the same magnet my template_list would include the 66 subjecs with pain and 66 randomly selected control patients without pain.
However, since the scans were collected using different magnets I want to create a template_list that includes possible effects from the different scanners.
Here is a list of the distribution of the subjects across the scanners.
Scanner ID Total each scanner #pain #control
1 30 7 23
3 46 13 33
4 47 12 35
5 39 15 24
6 9 2 7
7 20 7 13
8 50 10 40
241 66 175
Given the 2 pain subjects at scanner 6 I thought the most straightforward approach would be to randomly pick 2 pain patients from each scanner and 2 control patients from each scanner for a total 14 subjects in the template_list file. However, this approach seemed to pick too few subjects.
My next approach was to pick a number from each scanner proportional to the total number of subjects from each scanner. Using this approach I created a template_list with this proportional distribution of subject
Scanner ID total number of patients template_list
1 30 6 pain 6 control
3 46 10 pain 10 control
4 47 10 pain 10 control
5 39 8 pain 8 control
6 9 2 pain 2 control
7 20 4 pain 4 control
8 50 10 pain 10 control
Colm Connolly commented “my concern with the approach you've taken is that some magnets are more represented than others. If any of those magnets tend to over- or under- estimate the amount of GM in the brain you would bias the template in such a way. It might be better to drop the magnet with 2 pain & 2 control and choose 4 (the next smallest n) from each group and magnet and include those randomly chosen subjects in the template. I think you could include the subjects from magnet 6 in the VBM analysis even though there are no subjects from that magnet included in the template. I think that you should definitely ask for advice on this one on the FSL mailing list."
My question for you is do you think I should make my template_list file the way Colm suggests, or do you have another recommendation for making the template_list file for this analysis?
Thank you very much for your input!
John Keltner
|