MRI Slice image reconstruction

Hello all,

I have an MRI K-Space data 320 x 320 x 256 x 8 (4D complex double) from < mridata.org > . The data represents 320 x 320 K-Space of 256 slices from 8 channels. I am trying to reconstruct images for each slice. Here is the matlab code I tried:

    % Version - 1
    kspacedata= kspacefile.kdata;
    imRef = ifftshift(ifftn(kspacedata));
    imSOS    = squeeze(sqrt(sum( abs(imRef).^2, 4))); % sum-of-squares to combine all channels

    % Version - 2
    kspaceSlicedata = kspacedata(:,:,100,:);
    imSliceRef = ifftshift(ifftn(kspaceSlicedata));
    imSliceSOS    = squeeze(sqrt(sum(abs(imSliceRef).^2, 4))); % sum-of-squares to combine all channels

    % Plotting
    figure;
    subplot(1,2,1); imagesc(imSOS(:,:,100));title('Version - 1');axis image; colormap(gray);
    subplot(1,2,2); imagesc(imSliceSOS);title('Version - 2');axis image; colormap(gray);

In Version - 1, I take inverse transform on entire kspace data and plot image data corresponding to slice 100. In Version - 2, I take kspace data corresponding to slice 100 and do inverse transform on this kspace and plot the image data. Here is the output images I obtained.

I thought both versions will reconstruct the same way. But images appeared different. How can I take kspace data corresponding to one slice and perform inverse transform to reconstruct slice image?

Thanks.

Answer 1

[zoey yang]

the version 1 is 3D FFT, and version 2 is 2D FFT. It is about the way you acquired data. version 1: encode the entire 3D space and acquire the data to fill the k space. version 2: encode the slice and acquire the slice data then you encode the next slice and acquire data. That's the difference.

Answer 2

[Whae five]

https://www.mathworks.com/help/images/examples/exploring-slices-from-a-3-dimensional-mri-data-set.html

Answer 3

[FELIPE COSTA]

I tested here with another dataset and I was having the same problem. The function ifftn can be used for 3 or more dimensions. So it worked in your first version. Your second version is a 2D, so you must use ifft2 instead.