Using NaN in indexing a matrix

Hi all,

I'm writing a function that defines a sphere (center and radius are part of the inputs of the function), takes an fMRI scan file then finds the average activation value of all the voxels within that sphere. Below is what I do:

%Use make_sphere function to make a sphere with given s (grid size) and r (radius):

my_sphere = make_sphere(s,r)

%center of sphere:

center = [i, j, k]; 
my_sphere = bsxfun(@plus, my_sphere, center); %create a 3D matrix for sphere

%Access the scan file

Data2Read=fullfile(datadir,fname);

a=spm_vol(Data2Read);

v=spm_read_vols(a);

%find the mean activation of voxels within the sphere:

theSum = 0;

count = 0;

for h = 1:size(my_sphere, 2)

    theSum = theSum + v(my_sphere(h,1), my_sphere(h,2), my_sphere(h,3));

    count = count + 1;

end

MeanActivation = theSum / count

Since it is possible that part of the sphere may fall outside the brain (i.e. outside of the space defined in the scan file), I imagine values in this part of the sphere would be NAN. Thus, when taking the average, I could perhaps use isnan() to exclude these NANs from the average. However, whenever I encounter such scenarios, I only get an error message (Example: ??? Attempted to access v(11,77,63); index out of bounds because size(v)=[72,72,33].). Could anyone help?

Answer 1

[Image Analyst]

No, don't do it that way. See the last chunk of code in this section of the FAQ for how to create a 3D mask of a sphere that you can use to get your mean: http://matlab.wikia.com/wiki/FAQ#How_do_I_create_a_circle.3F

meanActivation = mean(your3DImage(sphereVoxels));

Like I said, you get sphereVoxels from the FAQ, and in there is where you specify the center and radius.