 |
 |
Hi Lau On Dec 4 2019, at 12:08 pm, laust vind knudsen <[log in to unmask]> wrote: > > Hello FSL experts! > > I´m a rookie and currently running structural analysis on some practice data, which consist of 26 Alzheimers and 7 Parkinson´s patients. I have performed FIRST, with vertex analysis and volumetric analysis (corrected for Vscaling) and VBM, and during these procedures certain questions have come up, which I haven’t been able to find answesr to on the JISCMAIL webside. > 1. When I create my design in the GLM_GUI I get this warning; > [“Problem with processing the model: Warning - design matrix uses different groups (for different variances), but these do not contain “separable” Evs for the different groups (it is necessary that, for each EV only one of the groups has non-zero values) This massage can be ignored if you are intending to use the groups file to define exchangeability block for randomize”] > - I´m pretty sure my design is good and that I corrected for age in the proper way, but I just wanted your judgement on whether I need to change it or not (The design is attached to the mail) > > > > The "Group" column is used to define separate variance groups for FEAT, or separate exchangeability blocks for randomise. Unless you have specific reasons to, I would suggest setting the "Group" column to 1 for all of your subjects. Otherwise, as the warning suggests, you would usually want to split your age EV into two EVs, one per group, with zeros for the out-of-group rows. > 2. In relation to this I get this message then I run randomise; > “Warning: Data demeaning selected, but at least one design column has non-zero mean - therefore invoking automatic demeaning of design matrix” > - I`m using the “-D” option for demeaning, but don’t understand what the problem is? > (randomise -i diff_AD_PD_L_Hipp_MNI.nii.gz -m diff_AD_PD_L_Hipp_MNI_mask.nii.gz -o AD_PD_L_Hipp_randomise_MNI -d AD_PD.mat -t AD_PD.con -f AD_PD.fts -D --T2) My guess is that this warning is related to the fact that you have defined separate exchangeability blocks in the Group column. > 3. in my FIRST analysis I used already brain-extracted image using the -b options. But it´s mention that it´s discouraged to use already brain-extracted images, how come? The reason I do it, is there is some variability in the parameters that works best for my subjects, therefore I think its preferable to execute brain-extraction beforehand. But what are your thought and reason for saying otherwise? Whole-head images are preferred, because they will (usually) allow for a better registration to the MNI152 template. FIRST works in the sub-cortical regions, so having a precise brain-extraction is not particularly important. > 4. In the Vertex analysis, if I want to make those cool images of a certain sub-cortical structure (mask) with the differences overlaid on it (corrp image) is it still preferable to use first3Dview. I ask because I’m having trouble with the program. You can use whatever image/data viewer you want. FSLeyes is capable of producing similar scenes to first3Dview, as outlined here: https://git.fmrib.ox.ac.uk/snippets/12 > 5. In VBM brain/head size is corrected for via transformation to standard space.I´m also using MNI space in vertex analysis, where I use "–useReconMNI" and "–UseRigidAlign", which I read was advantageous as it corrects for absolute and relative pose and head/brain size. > But when talking vertex analysis, would it be better to do the analysis in native space and correct for head size via Vscaling instead? > Should I add head/brain size as a covariate even though it´s in MNI-space and therefore already is corrected for? It really on the specific question you are asking. Are you testing for differences in absolute size? Or are you looking for differences in shape or pose/orientation? > 6 (LAST QUESTION I PROMISE) In relation to VBM, if i have had my pictures through the “fsl_anat” script, thereby removing neck (among other things) which means I should take the -a option, right? (e.g. fslvbm_2_template -a) The -a option tells the VBM pipeline to use linear, rather than nonlinear registration; this is typically not a good idea for VBM, as nonlinear registration will better align the images from all of your subjects to the template. Paul > Hope to hear from you and sorry for the many questions. > Kind regards Lau To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 ######################################################################## To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1