[Marsbar-users] Beta Value Extraction with marsbar 0.42

[Esther D. <es...@gw...>]

Dear Matthew Brett,

 

I have an important question regarding "beta value extraction" with marsbar
0.42.

I use the following code to extract parameter estimates:

 

        D = mardo(spm_name);

        Y = get_marsy(roi, D, 'mean');

        E = estimate(D,Y);

        xCon = get_contrasts(D);

        E=set_contrasts(E, xCon);

        b = betas(E);

        marsS = compute_contrasts(E, 1:length(xCon));

        cont = marsS.con; 

        Tval = marsS.stat; 

        pval = marsS.P; 

        pcorr = marsS.Pc; 

        f = length(Tval);

        Roinr = zeros(1,f)'+i;      

        connr=zeros(1,f)';

                for k = 1:length(connr)            

           connr(k,1)=k;

        end 

 

My data have a voxelsize of 3x3x3 mm^3 and I extracted the parameter
estimates from spherical ROIs centered at peak coordinates from previous
neuroimaging studies. However, I encountered the problem that some previous
studies used a different voxel resultion than me (e.g. 4x4x4 mm^3). 

I therefore tried to do the following, which I would like to outline using
an example:

 

I wanted to extract the subject-specific beta value for "CONTRAST 1" from a
spherical ROI in the left ventral striatum. A previous study reported a
maximum at "-12 12 -4". So I used this maximum as center for a spherical ROI
with a radius of 4 mm. 

(1.) Center at -12 12 -4 with radius of 4 mm^3 ---> This yielded a beta
value of 1.520 for subject 1.

 

 

Still, I was unsure whether with ROI was possible with my voxelsize of 3x3x3
mm^3 or whether MARSBAR would interpolate the result by using (multiples of)
my voxelsize for center coordinates and/ or radius. So I repeated the
procedure with a converted center-coordinate that existed in my voxelspace
(i.e., "-12 12 -3") with a radius of 3mm (as a multiple of voxelsize). 

(2. ) Center at -12 12 -3 with radius of 3mm^3 ---> This yielded a beta
value of 1.494 for subject 1. 

 

 

So far so good. But I did not stop there, but also performed two additional
ROI extractions:

 

(3.) Center at -12 12 -4 with radius of 3mm^3 ---> This led to a beta value
of 1.667 for subject 1.

 

(4.) Center at -12 12 -3 with radius of 4mm^3 ---> This led to a beta value
of 1.494 for subject 1, which is identical to the one derived from a sphere
with the 3 mm radius (see (2.))!!!

 

 

But how is this possible? And why was the resulting beta only identical for
the ROIs described under (2.) and (4.)?

 

If MARSBAR indeed interpolates by using the original voxelsize of the data,
than I would have expected all four extractions to lead to the same result. 

 

If MARSBAR only interpolates the radius to a multiple of voxelsize, but not
the center coordinate than both (2.) and (4.) and also both (1.) and (3.)
should have been identical.

 

 

I hope that you can help me to resolve this problem. Is there anything in
the code that I have missed?

 

 

Thanks in advance!

 

 

Best regards,

 

Esther Diekhof