help > Subject space analysis - normalise beta maps?
Showing 1-3 of 3 posts
Apr 6, 2017 10:04 AM | Rob McCutcheon - Institute of Psychiatry
Subject space analysis - normalise beta maps?
Dear Alfonso and others,
I am attempting a resting state analysis in which I wish to perform the first level analysis in subject space (I have previously performed it in MNI space, and am wanting to check that the results remain the same with this method).
In my preprocessing I center the images, realign and unwarp the functional, slice time correct, coregister, segment the structural, and use ART for outlier detection. I then use the i_y*.nii files generated during the segmentation to move my ROIs (originally in MNI) to subject space, which I do for each subject
My questions regards how to perform the second level analysis - how do I normalise/smooth the beta maps to MNI space for this step? If I just go to the second level analysis tab as things stand my results currently do not show the connectivity patterns I expect (e.g. just a few random scattered voxels for the second level results).
Many thanks,
Rob
I am attempting a resting state analysis in which I wish to perform the first level analysis in subject space (I have previously performed it in MNI space, and am wanting to check that the results remain the same with this method).
In my preprocessing I center the images, realign and unwarp the functional, slice time correct, coregister, segment the structural, and use ART for outlier detection. I then use the i_y*.nii files generated during the segmentation to move my ROIs (originally in MNI) to subject space, which I do for each subject
My questions regards how to perform the second level analysis - how do I normalise/smooth the beta maps to MNI space for this step? If I just go to the second level analysis tab as things stand my results currently do not show the connectivity patterns I expect (e.g. just a few random scattered voxels for the second level results).
Many thanks,
Rob
Apr 10, 2017 02:04 PM | Alfonso Nieto-Castanon - Boston University
RE: Subject space analysis - normalise beta maps?
Dear Rob,
That is an interesting question. Currently there is no simple way to do this within CONN (if you are doing analyses in subject-space then you can perform ROI-to-ROI second-level analyses normally, since those ROIs are also defined subject-space, but second-level seed-to-voxel or voxel-to-voxel analyses will be meaningless, since there is no appropriate inter-subject coregistration within voxels; for that you need your data normalized first). You would have to take your BETA_*.img files generated by CONN's first-level analysis step and manually re-warp them to MNI space before entering them into your second-level analyses. For example, you could use SPM's 'normalize (write)' option and use your original y_*.nii files deformation fields (separately for each subject) to generate the MNI-space beta maps. This is a bit tricky to get right, though, since you need to make sure that your 'subject-space' was actually the same as that used as source in your original normalization step (e.g. the structural/functional data may have been centered before running normalization, in which case you would also need to apply the same centering transformation to your BETA images; or your subject-space pipeline may have coregistered your functional to your structural data while your MNI pipeline may have performed normalization on the original functional data -i.e. pre-coregistration). If you let me know the exact procedure that was used to preprocess your data in both cases I should be able to give you a more precise answer. That said, I do not see any reason why the results from this procedure should be any different from the results of your original MNI-space pipeline (as far as I can tell the results should be very similar if not identical, the only difference being perhaps due to expectedly-very-minor resampling/interpolation effects arising from resampling your ROIs+BETA maps in one case vs. resampling your BOLD-data in the other)
That is an interesting question. Currently there is no simple way to do this within CONN (if you are doing analyses in subject-space then you can perform ROI-to-ROI second-level analyses normally, since those ROIs are also defined subject-space, but second-level seed-to-voxel or voxel-to-voxel analyses will be meaningless, since there is no appropriate inter-subject coregistration within voxels; for that you need your data normalized first). You would have to take your BETA_*.img files generated by CONN's first-level analysis step and manually re-warp them to MNI space before entering them into your second-level analyses. For example, you could use SPM's 'normalize (write)' option and use your original y_*.nii files deformation fields (separately for each subject) to generate the MNI-space beta maps. This is a bit tricky to get right, though, since you need to make sure that your 'subject-space' was actually the same as that used as source in your original normalization step (e.g. the structural/functional data may have been centered before running normalization, in which case you would also need to apply the same centering transformation to your BETA images; or your subject-space pipeline may have coregistered your functional to your structural data while your MNI pipeline may have performed normalization on the original functional data -i.e. pre-coregistration). If you let me know the exact procedure that was used to preprocess your data in both cases I should be able to give you a more precise answer. That said, I do not see any reason why the results from this procedure should be any different from the results of your original MNI-space pipeline (as far as I can tell the results should be very similar if not identical, the only difference being perhaps due to expectedly-very-minor resampling/interpolation effects arising from resampling your ROIs+BETA maps in one case vs. resampling your BOLD-data in the other)
May 10, 2018 05:05 PM | Daniel Berge - Hospital del Mar Medical Research Institute (IMIM)
RE: Subject space analysis - normalise beta maps?
Hei,
Maybe a stupid question, but, would it be possible to extract the ROI beta time series in subject space, but find out correlations with beta-time series in mni space? (similarly to what actually conn does extracting ROI beta-time series from unsmoothed images and correlating with time-series from the smoothed images).
Thanks
Maybe a stupid question, but, would it be possible to extract the ROI beta time series in subject space, but find out correlations with beta-time series in mni space? (similarly to what actually conn does extracting ROI beta-time series from unsmoothed images and correlating with time-series from the smoothed images).
Thanks