Not Working for standard Spin-Echo EPI

[wenshangwang]

Dear Mengye

I am trying to handle the N/2 ghosting artifact in my Spin-Echo EPI image after reconstructed from k-space data. I put the data in the desired format (kx, ky, slice, coil) and run your code:

Where:

is my data with 128 readout samples, 64 PE lines, 26 slices and 44 channels. Which is one echo of the SE EPI.

What suggestion would you have for me to try. Thanks you very much

Wenshang

[mylyu]

Sorry for the late reply. Was your data parallel image accelerated? If so, the entropy based correction method won't work well. In such case, I usually acquire a fully sampled reference scan (e.g. multishot EPI, 3-echo navigator) to run the entropy based correction and apply the obtained parameters to the actual accelerated data. Alternatively, I recall there are some low rank based correction methods that can work with accelerated data.

[wenshangwang]

Sorry, the project was sideline for a bit so I did not reply earlier. Thanks for your reply.

My data was fully sampled SE-epi data. It is parallel image accelerated from 44 coils, However I am a little confused. As "nyquist_ghost_correction" function would take multi coil data, isn't this imply the function should already works with parallel imaging data?

Please correct me if you think there are anything wrong with my statement. Looking forward for your reply.

Best
Wenshang

[mylyu]

Sorry, the project was sideline for a bit so I did not reply earlier. Thanks for your reply.

My data was fully sampled SE-epi data. It is parallel image accelerated from 44 coils, However I am a little confused. As "nyquist_ghost_correction" function would take multi coil data, isn't this imply the function should already works with parallel imaging data?

Please correct me if you think there are anything wrong with my statement. Looking forward for your reply.

Best Wenshang

Parallel imaging accelerated means not fully sampled. The "nyquist_ghost_correction" function does take multi coil data, but the data should be fully sampled (i.e., not parallel imaging accelerated).