Skull-strip partial field-of-view T2 image?


I am trying to skull strip a T2 image acquired using a 7T OP coil (occipital parietal). The image has full inferior-superior and right-left coverage. The anterior-posterior coverage is only the back half of the brain with a signal gradient running in the AP direction.

I have tried FSL’s bet with limited success. I cannot find an optimal fractional intensity value that results in a complete skull strip without also removing brain. The gradient option appears to be hard coded for the IS direction. (I haven’t tested reorienting the image to swap the IS and AP directions to take advantage of the gradient option.)

AFNI’s 3dSkullStrip defaults performed significantly worse than FSL’s bet defaults. I haven’t tested all of the options for 3dSkullStrip. I was honestly more concerned about the skull-stripped image no longer being aligned with the original input image.

I’m not super familiar with ANTs skull stripping to know whether or not it is flexible enough to accommodate partial FOV images. Also, surface-based techniques aren’t really an option, as the study is interested in cerebellar and brainstem nuclei.

Is there a skull-stripping algorithm that is designed to take in partial FOV images, or am I on a fool’s errand?



You might want to see if HD-BET works - if you have a CUDA capable GPU this is super fast (< 1 minute).

You may also want to try CAT12. You may want to manually set the origin to be near the anterior commissure (using SPM’s Display option). By default, the origin will be the iso-center of your MRI scanner, which may be a poor starting estimate.

Hi Chris,

Thanks for the quick response. I will look into both of your suggestions.