Hi,

It is a common practice in MRI to zero fill the k-space array before performing the Fourier transform in image reconstruction. For historical algorithmic reasons, powers of two are often selected. In the cases you note, the default behavior of the GE system is to zero fill the acquisition to 256. There is relatively old literature which suggests that zero filling, although it offers no new information, yields images with improved aesthetics for diagnostic reading up to an increase in "resolution" to a factor of about 2. There are a few ways to get to the 128 matrix you're expecting:

1) There are control variables on a research enabled system to set the reconstruction matrix size. I don't have the source code in front of me, but if you ask this question on the GE Healthcare research collaboration portal (https://collaborate.mr.gehealthcare.com/), a very thorough answer will be quickly offered (assuming that it is not already answered in the forums!).

2) You can collect the raw k-space "p-file" on a pre-DV26 system (the format has changed with DV26, I think it's now an HDF5 file named differently) and reconstruct the images with either your own software, or using the outstanding Orchestra SDK which is offered through GE Healthcare's online research collaborative portal.

3) With a spin echo or partial Fourier acquisition (generally employed with DTI) and with coils combined through a sum of squares method (generally employed on MR systems without SENSE or ASSET acceleration), it is a very good approximation to the non-zero filling reconstruction to inverse Fourier transform the 256 x 256 pixel array, remove the symmetric frame of 64 zeroes, and forward Fourier transform to yield the dimensionality you expect. I ask my students to mathematically prove this statement in the class I teach. This is likely the only applicable solution to apply to data you have already acquired.

Kind regards,
Andrew Nencka