Thank you for this very helpful information. I realized that normally I had been using smwc1* images, and along the way, I had lost sight of the fact that these are gray matter images, rather than whole brain. Of course, I do realize that an inverse correlation can be read in either direction; I am not sure why I became confused.Now I am working on an analysis of gray matter density.I have chosen Preserve Concentrations in the dartel to mni step, and that has produced swc* images.Here is my new question:Is it possible to extract values for density of gray matter?
When working with 'gray matter volume', I used get_totals.m for that
(as well as for white and CSF, to calculate tiv).However, that algorithm produces 'volume in ml'.My intention is to use gray matter density in a behavioral analysis.
(I realize I do not need to calculate tiv when working with density because we do not control for tiv.)Thank you.JulieOn Tue, Jun 9, 2015 at 2:45 PM, MCLAREN, Donald <[log in to unmask]> wrote:See inline responses below.On Mon, Jun 8, 2015 at 10:35 PM, Julie Morgan <[log in to unmask]> wrote:Then, as covariates: score, age, tiv.(from which I had derived the gray matter volume information).Instead of inputting 'scans', I should input each person's c1 imageTherefore, is this what you mean?Yes, I do want to find regions in the structural gray matter (volume) that are inversely correlated with scores. Specifically, those regions of gray matter that are less involved as scores increase.Hello Donald and List,Thank you for coming to my assistance.You should enter the smwc1* images - smoothed, modulated, normalized gray matter images.Then, for contrasts, I notice that SPM automatically refers to column one as 'mean'; is that a reference to the scans / c1 images, or to the intercept?I presume that none of the covariates are set to interact with any factor...It its the mean or intercept of the scans/smwc1 images. It depends on how you entered the covariates. For more details see:I suspect that I am actually getting regions of gray matter that INCREASE as scores decrease (which is, unfortunately, not what I am querying).Is that correct? This confuses me because I think that -1 1 implies a t-contrast (subtraction rather than correlation....), but if I use 'zero' for scores, then the scores will be set aside from the analysis. Or, if my contrast is 0 -1 0 0mean score age tivIf 'mean' is actually the column of c1 / scan image information, then perhaps the contrast is:
-1 1 0 0This contrast doesn't make any sense. You are testing whether the slope is greater the than mean gray matter value. What you want to do with VBM and any other analysis is to start of with your null hypothesis.Ho: slope of score=0 [e.g. as GM goes up, the score goes down AND as GM does down, the score goes UP -- these are actually the same].Now, you'd make the null hypothesis equal to zero, if it already not equal to zero (e.g. Ho: A=B becomes A-B=0).Now assign the coefficients of the null hypothesis to the correct columns, all other columns get set to 0.The contrast you want to test is: [0 -1 0 0] or [0 1 0 0]The -1 contrast will test for an inverse/negative relationship or GM and score; while the 1 contrast tests for a positive relationship between GM and score.Hope this helps.Or perhaps what I should be doing is an F contrast?mean score age tiv
-1 0 0 0
0 1 0 0Don't use the F-test. You don't care where GM is different than 0 OR the score is positively related.JulieThank you.On Mon, Jun 8, 2015 at 3:16 PM, MCLAREN, Donald <[log in to unmask]> wrote:Julie,If you want to look at neural activation, then you to do an fMRI study, not at VBM study. Putting that aside, it seems that you hypothesis is that want to find areas where grey matter volume is negatively associated with your behavioral score, controlling for age and TIV.Multiple linear regression model:(1) Grey matter maps as the imaging dependent variable.(2) Independent variables are: score, age, and TIV.If you have a different hypothesis, please explain it in more detailBest Regards, Donald McLaren
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Office: (773) 406-2464
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.On Sun, Jun 7, 2015 at 11:34 PM, Julie Morgan <[log in to unmask]> wrote:
Dear SPM experts,
I conducted a behavioral analysis that showed a significant inverse correlation between scores and gray matter (volume) and a significant positive correlation between scores and age. I had covaried out total intracranial volume (tiv).
There is multicollinearity between age and gray matter. However, I can predict scores (regression) separately from age and from gray matter.
Now I would like to run an analysis in VBM that would identify neural activation associated with decreased gray matter volume as scores increase.
I tried Multiple Regression, entering all the scans in one “go”, then scores, gray matter, and tiv (one column each, no interactions or centering for any of these.) I set the absolute threshold to 0.2, said ‘yes’ for an implicit mask, and omitted global calculation/normalisation.
The design matrix has four columns:
mean (automatic name given by SPM) scores gray tiv
Have I set this up correctly?
How do I test to show neural activation associated with decreased gray matter volume as scores increase?
I searched the helplist and found John Ashburner’s email of June 27, 2003, and well as several other emails about multiple regression using VBM. In that email it is suggested that we should first create groups as a condition using ones and zeros.
However, when I created an F-contrast with
for each participant in agegroup1 1 0 0 (several rows)
for each participant in agegroup2 0 1 0 (several rows)
for each participant in agegroup3 0 0 1 (several rows)
the multiple regression batch interface would not accept that ‘vector’. (A vector is one row or column, not an F contrast.....).
I can enter 3 groups (the scans themselves) using a factorial design specification but then I am not doing multiple regression.
In multiple regression, I could ‘create three groups’ if I entered group (or age, same thing) and specified ‘interaction with factor 1’ but what would factor 1 be?
My goal is, still, to run an analysis in VBM that would identify neural activation associated with decreased gray matter volume as scores increase.
Could someone please advise me?