I have an MRI kspace from a siemens 1.5 T scanner with reduced scan percentage ( sampled truncation by not acquiring the most peripheral lines in kspace). The kspace plot in matlab look like below.

I can do zero filling in this kspace to the required resolution and reconstruct images.

enter image description here

Other than zero filling, is there another method which can estimate missing lines and reconstruction images.


I would say no, there isn't. This is unmeasured information and you cannot compute it without additional information (ie some additionally sampled data, and maybe some compressed sensing algorithm). However, truncating an image like that in order to save measurement time is quite unusual. You would be better off to sample one side full and skip twice the number of lines on one side of $k$-space. You could then use additional information (in this case: the object is real and should be shimmed well, ie you know that the image data should be real). The $k$-space of a real image is Hermitian, and you could use the sampled information of the one side outer part to use the Hermitian symmetry to fill the missing lines. Different algorithms for this exist.

  • $\begingroup$ Thanks @M529. We collect MR data of a Siemens 1.5T machine from a hospital in our city. I usually collect data for T1 or T2 TSE sequences of brain scan. Everytime I get data with sampled truncation from the scanner. But your comment says it is an unusual way of reducing scan time. Is this the default scan settings of Siemens machine? If I collect data for a haste sequence, I am getting partial fourier data. Otherwise I get data with sampled truncation. $\endgroup$
    – Thara
    May 23 '19 at 10:31
  • 1
    $\begingroup$ TSE is a bit special, since the number of phase encoding steps that you would like to have (ie due to resolution and FOV) and the Turbo Factor get a bit entangled here. So it could be that the data you expect (ie due to resolution and FOV) is a bit different from what you get (since Turbo Factor is taken into account). Are you also sure that you have set the resolution in PE direction to 100%? $\endgroup$
    – M529
    May 23 '19 at 18:07
  • $\begingroup$ Thanks for the answer. Since operator is scanning and acquiring data, I am not sure about resolution in PE direction. Hopefully, I can check it next time. So may be with SE sequences, I will getting fully sampled kspace, right? $\endgroup$
    – Thara
    May 24 '19 at 4:42
  • $\begingroup$ I assume so. Depending on your applicaiton, it might be the best to use simple sequences without any acceleration techniques for first approaching the subject. $\endgroup$
    – M529
    May 26 '19 at 17:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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