Dear UKB neuroimaging team,
We are analyzing T1 scans of UKB's stroke patients, running an Automated Lesion Identification tool (implemented in SPM) to create binary lesion masks, which are then used for a lesion overlap analysis. I wanted to ask about two things:
1. As we need to use a different segmentation and spatial normalization procedure than the one used in the UKB pipeline (as this was presumably not optimised for patients with stroke lesions), is the the most suitable file to use the "T1" file, that is, with only the gradient distortion correction (and defacing) applied?
If I understand correctly, at that stage of pre-processing, there will be the issue of internal carotid arteries being brighter in the UKB scans (than in the MNI template) and therefore subsequent registration to MNI space may not work optimally. Would you advise us to try and implement this in our pipeline?
2. We would ideally want to include also the scans which were assessed as "unusable" (due to having gross lesions) in your quality control pipeline, and therefore not processed. As they are of interest to us, I was wondering if we could use the "T1_orig_defaced" files to run our segmentation and spatial normalization pipeline. In addition to the issue with brighter internal carotid arteries mentioned above, the "T1_orig_defaced" files don't have gradient distortion correction applied.
As you say in the paper "Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank", gradient distortion correction substantially improves alignment. However, doing it would require getting the proprietary Siemens file with the gradient coefficients, which may not be entirely straightforward to do. And from the images presented in the above-mentioned paper, it seems that the difference may be only a few voxels - I am not quite able to tell if this would be a major issue for the purposes of our study, considering that we apply spatial smoothing (and are only interested in lesions greater than 400 or even 800 voxels).
Would you be able to advise if you think that applying gradient distortion correction is advisable if we want to use the "unusable" scans for the purposes of our study? And if yes, could you advise how difficult this would be? (The issue is that if it should take a substantial amount of time and effort for only a modest improvement in alignment, it just may not be worth it for the purposes of our study...)
Many thanks and best wishes,
Karel
---
Karel Kieslich
PhD student
UCL Institute of Cognitive Neuroscience
########################################################################
To unsubscribe from the UKB-NEUROIMAGING list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=UKB-NEUROIMAGING&A=1
This message was issued to members of www.jiscmail.ac.uk/UKB-NEUROIMAGING, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|