suit isolation issue

[yenwchen]

Hi!

We ran into a problem that the cropped image from 'suit_isolate_seg' step didn't fully cover the cerebellum for some of our participants' T1w. We adjusted the bounding box using the 'bb' command: [-76 76; -126 -50; -40 40].

The issue we had was that the -y = -126 seemed to be the bound of the most posterior that could go. Was this the posterior bound of the MNI template?

We believe that our participants (age between 8 - 17) T1w are not larger to the point that go beyond the MNI template bound. We suspect that, for some reason, the participant's T1w was being ‘shifted’ posterior to outside of the MNI boundary during the process. Would you kindly let us know what could be the potential issue underlying this behavior?

For your reference, the T1w image was acquired with MP2RAGE and we combined two T1w with different inversion times into a uniformed T1w. The T1w is LPI and we manually set the origin at AC using SPM.

We greatly appreciate your insights on this matter. Thank you.

Best regards, Yen-Wen

[jdiedrichsen]

No - it just looks like that the intial SPM normalization to MNI failed on this scan. If the AC is set correctly, I'm not sure why this is - possibly the well-preserved contrast in the cheeks - or the overall smaller head size. If you look in the isolation code, you probably can figure out where the inital MNI normalization happens (before cropping) and force the affine alignment manually to something better. It's a hack, but we are wroking on a complete rewriting of the isolation step - so by September we hopefully have a better version out.

[yenwchen]

Hi Dr. Diedrichsen! Thank you for your response. I went through the script and tested where the issue came from. The segmentation went fine. Please find the screenshot of the segmentation output. As you mentioned, it's the cropped step. Our participants are on the younger side (8 - 17 years old, and the ones with issues are between 9 - 11) and their sform values are larger. The workaround we came up with was that to manually specify the bounds (in voxel location) for "spmj_crop_vol". The output cropped images look fine. Please find the attached screenshot. Please don't hesitate to let us know if this adjustment doesn't make sense. We look forward to the updated version for isolation! Once again, thank you for your time and suggestions.