 |
 |
Dear Mark Thank you for your help. Your answers are of great help to me, and all my quesries are now clarified. Sorry for bothering you again with additional 2 questions. Q1) I guess *_pval.vtk represents uncorrected p values. Am I right? And I would like to display only regional shape changes (atrophy) that survived FDR-corrected P < 0.05. How can I do this? I extensively searched the FSL mailing lists, but found no thread. Q2) From my little experience, I'm afraid that shape analysis (e.g,, FIRST, FreeSurfer) is more sensitive in evaluating regional atrophy in subcortical GM structures than VBM-style analysis (e.g., VBM8, DARTEL, FSL-VBM). If so, is there a research paper that showed shape analysis using vertex analysis is superior than VBM-style analysis in subcortical GM examinaton ? Thanks in advance for your help. Best wishes, Min J. Kim Date: Mon, 7 May 2012 14:53:54 +1000 From: [log in to unmask] Subject: Re: [FSL] Questions about FIRST-vertex analysis To: [log in to unmask] Dear Kim, I'm currently working on shape analysis of thalamus between patient and control groups using FIRST vertex analysis. Using fsl glm, I set up the following design matrix (please see the attached figure): Group column: automatically ignored by vertex analysis? EV1: control 0, patient 1 EV2: age EV3: sex (male 1, female 2) Q1) I want to compare thalamic shapes between the groups with age and sex as covariates-of-no interest. Is above design matrix right? If not, please correct me. FIRST will demean the EVs for you and so this design matrix looks OK. Q2) After running the vertex analysis as described in FSL website, I found 3 vtk files for each thalamus (e.g., L_Thal1.vtk, L_Thal2.vtk, L_Thal3.vtk) in the shape_analysis directory. I guess L_Thal1.vtk represents between-group difference of left thalamic shape after *controlling for age and sex*. Am I right? However I cannot figure out what the L_Thal2.vtk and L_Thal3.vtk mean? Are those 2 vtk files the results of correlation of each covariate (age, sex)? FIRST creates a simple contrast for each EV. So the results numbered 1 are for the contrast [1 0 0]and those numbered 2 are for [0 1 0] and those numbered 3 are for [0 0 1]. In your design it is really thefirst result (effectively the contrast [1 0 0]) that is of interest, as it looks at the group difference with theother EVs as regressors of no interest (or confounds). Q3) After performing FDR correction using surface_fdr command, I got significant results (L_Thal1: 0.039, R_Thal1: 0.034). When I click the add-mesh button and choose either L_Thal1._Fthresh.vtk or L_Thal1_pval.vtk files after making 3D viewer in fslview, I found signigicant regional differences of both thalami. However, there seems *No* little arrows on the thalamic surface, which indicate direction of changes. Is there something wrong with my FSL? I'm using FSL 4.1.9 on CentOS 5.8. Try modifying the size of the arrows in the pop-up properties GUI. If you are notsure about this look at the FSLView documentation or the Segmentation practicalin the FSL course. Q4) In addition, I want to know which thalamic region is positively or negatively correlated with disease duration or disease severity score in the patient group. How can I set up the design matrix for this purpose? If you add the disease duration (or severity score) as an EV in the design matrixthen you will get the correlation with the other EVs treated as confounds. If you donot want to have confounds then just create a design matrix with only one EV.You also need to do this only for the patients (so make a new concatenatedbvars file that only includes the patients, and make sure that the design matrixonly includes rows for these patients). Apologies for the multilple questions. Thank you in advance for your help. No worries.All the best, Mark Best wishes, Kim <Screenshot-General Linear Model.png>