I have been dealing with an issue during preprocessing in CONN in which part of the skull is still present in my structural image after skull-stripping for one of my subjects. For my pipline I used indirect segmentation and normalization. There were no errors shown during the normalization step. When looking at the segmentation for the CSF if appears to inlcude areas of the skull. However the gray and white matter segmentation look normal.
I tried all the options for segmentation and normalization (direct/indirect, seperately, etc.) but this issue still comes up. I have also tried to resolve this by conducting skull-stripping outside of CONN (I used BET in FSL), which resulted in a properly skull-stripped image. However, when I tried applying the normalization step to the skull-stripped image with my new image, it did not do much to improve the alignment with the MNI template.
Do you have any ideas on how I might fix this?
Hi,
Incorrect segmentation masks typically means that the normalization procedure did not converge correctly. Have you tried manually re-positioning the original/raw structural image (just using manual rotations and translations) so that it is closer in position and orientation to the template and then re-running normalization again to see if that helps the procedure converge and produce more accurate segmentation outputs?
Best
Alfonso
Originally posted by e ble1:
I have been dealing with an issue during preprocessing in CONN in which part of the skull is still present in my structural image after skull-stripping for one of my subjects. For my pipline I used indirect segmentation and normalization. There were no errors shown during the normalization step. When looking at the segmentation for the CSF if appears to inlcude areas of the skull. However the gray and white matter segmentation look normal.
I tried all the options for segmentation and normalization (direct/indirect, seperately, etc.) but this issue still comes up. I have also tried to resolve this by conducting skull-stripping outside of CONN (I used BET in FSL), which resulted in a properly skull-stripped image. However, when I tried applying the normalization step to the skull-stripped image with my new image, it did not do much to improve the alignment with the MNI template.
Do you have any ideas on how I might fix this?