How to Improve Preprocessing for PA-Direction Multi-Echo EPI?

, ,
1

Hello,

I’m reaching out for some advice regarding preprocessing of data acquired in the PA (posterior-to-anterior) phase encoding direction on a 3T scanner.

After preprocessing, I’ve noticed that the overall brain size appears reduced, particularly with noticeable signal loss or distortion in the frontal regions, compared to data acquired with the same sequence in the AP direction. It also seems that distortion correction and normalization may not be working properly.

Here are the details of my acquisition and preprocessing steps:

  • Sequence: Multi-echo EPI (TR = 1s; TE1 = 13.00 ms, TE2 = 31.26 ms, TE3 = 49.52 ms; FoV = 210 mm; multiband factor = 4)
  • Preprocessing pipeline:
    1. Outlier detection
    2. Slice timing correction (SPM12)
    3. Motion correction (FSL)
    4. Distortion correction (FSL)
    5. Denoising using TEDANA
    6. Coregistration and normalization (SPM12)
    7. Spatial smoothing (6mm FWHM, SPM12)

I would greatly appreciate any suggestions or insights—especially related to preprocessing of PA-direction EPI images or multi-echo sequences—that might help resolve the issues I’m encountering.

Below is an example of one participant’s image after spatial smoothing has been completed.

스크린샷 2025-05-26 오전 9.55.41
스크린샷 2025-05-26 오전 9.55.411451×822 151 KB

스크린샷 2025-05-26 오전 9.55.58
스크린샷 2025-05-26 오전 9.55.581451×822 156 KB

Thank you in advance for your help!

Best regard,
Eui-Jin

2

Hi Euigenie,

PA will definitely have different distortions and possibly different dropout than AP. It might vary by datasets & brains so I wouldn’t expect it to be consistently worse for PA.

Could you share a PA & AP image from the same brain so people here can see if the difference is plausible or a sign of some other issue?

If there is an issue, it could happen in several places.

  1. Acquisition
  2. Distortion correction
  3. Masking (unclear where this is happening in your pipeline). The tedana group recommends people provide a mask made in another pipeline & visually checked instead of using the auto-mask within tedana. (See: FAQ — tedana 25.0.1 documentation)

If you look at or share images from each of those phases, it might be possible to better see where the problem arises.

Hope this helps

Dan