I reconstruct images from MRI k-space using ifft and root-sum-of-squares method.

imRef = ifftshift(ifftshift(ifft(ifft( ifftshift(ifftshift( kspacedata,1),2),[],1),[],2),1),2); 
imSOS    = squeeze(sqrt(sum( abs(angle(imRef)).^2, 3))); 

k = abs(imSOS); 
disp(class(k)); %double

Class of image matrix k is shown as double.

Is the image matrix k an indexed image or intensity image?

The sample values from reconstructed image k is shown below.

4.2753    4.9807    4.5435    5.6548    6.1303    5.0229
3.3805    4.6260    5.1594    5.2692    4.1187    4.4885
5.8990    5.5275    4.3493    5.6182    6.7237    4.3071
6.4885    4.6861    4.4086    3.5034    5.2378    4.7466
6.1099    4.6995    4.1673    4.7408    3.8915    5.7531
5.4006    3.1289    5.6541    5.8782    4.6568    3.8166

By referrring http://in.mathworks.com/help/matlab/creating_plots/image-types.html, I learned that for indexed images, the data matrix is represented by integers which are indices to color map. And if it were intensity images, I feel the data range is somewhere between 0 and 1. If k represents an indexed image, how can I know the colormap associated with it?



2 Answers 2


FFT and IFFT are linear operators, and as such, the results only make a lot of sense in a linear intensity space, not if indexed into a non-linearly mapped space.


In MRI, there is no intrinsic "color map". Opposed to Computed Tomography, there is no analogy to the Hounsfield Units. Hence, you only have relative intensity information.

Commonly, if you reconstruct your images by yourself, you'll have double values in an arbitrary range. If you compare that to the DICOM data from the MR machine, you'll notice that the DICOM images are rescaled to integer values, lets say between 0 and 4095. The rescaling factor is just a number to fit the signals in a predefined range - again no additional quantitative information here.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.