Hello,

I was wondering how one would reconstruct the FA map from the diffusion and kurtosis tensor MAT files? I ran the DKE for several subjects but left the map interpolation on, and I was hoping reconstructing fa.nii would be faster than running the DKE again without map interpolation.

Thanks!
Christina

Hi Christina,

You just need to calculate the fa parameter for each column of the DT.mat matrix and then reconstruct the image. One potential solution if you have MATLAB is to download spm8 and then run the script below. Alternatively, you may reconstruct the diffusion tensor for each voxel; dt = [DT(1,i) DT(4,i) DT(5,i);DT(4,i) DT(2,i) DT(6,i);DT(5,i) DT(6,i) DT(3,i)];, and then get the eigenvalues and solve the conventional equation. Hope that helps...

Best,

Russell

%fn_img: file path 'template' image (.nii) (eg b0 3D or 4D)
%fn_dt: file path to DT.mat (.mat)
%fn_fa: file path to new fa image to write (.nii)

hdr = spm_vol(fn_img);
load(fn_dt)
idx = sqrt(sum(DT.^2))>0;
fa = zeros(1,prod(hdr(1).dim(1:3)));
fa(idx) = sqrt(3.*sum([DT(1:3,idx)-repmat(sum(DT(1:3,idx))./3,3,1);repmat(DT(4:6,idx),2,1)].^2)./...
sum([DT(1:3,idx);repmat(DT(4:6,idx),2,1)].^2)./2);
hdr(1).fname = fn_fa;
hdr(1).dt = [16 0];
spm_write_vol(hdr(1),reshape(fa,hdr(1).dim(1:3)));

Worked like a charm. Thanks!

Can you provide matrix definition similar to dt = [DT(1,i) DT(4,i) DT(5,i);DT(4,i) DT(2,i) DT(6,i);DT(5,i) DT(6,i) DT(3,i)];, for the kurtosis tensor (KT)? 

Thanks!
Kathryn

Hi Kathryn,
 
Sure.  Each column is given by
 
W(:,i) =  [W1111, W2222, W3333, W1112, W1113, W1222, W1333, W2223,W2333, W1122, W1133, W2233, W1123, W1223, W1233]^T, where W is the kurtosis tensor for the ith voxel.
 
Best,
 
Russell

Thanks!!