I am reposting this message, as I didn't get any answer so far: -------------------------------------------------------------------------------------
Dear Freesurfer experts I have run a group ROI analysis comparing cortical thickness and area between two groups of subject. I did this in two different ways: (1) using mri_segstats after transforming each subject's data to fsaverage space, AND (2) using mris_anatomical_stats after transforming the ROI from fsaverage space to the space of each subjects. I get very similar results for cortical thickness, but completely different results for area. Does anyone have any idea why that could be happening ? Details: I have an ROI that I obtained from an atlas in MNI space and projected to the fsaverage surface (into say ROI_lh.mgh file). The aim of the ROI analysis was to find whether there are differences in cortical thickness and area between two groups of subjects in this ROI. I tested this in two ways: Method 1: a) I transformed the cortical thickness and area data of each subject into fsaverage space mri_surf2surf --s subj --trgsubject fsaverage --hemi lh --sval area --tval subj_area_fsaverage.mgh mri_surf2surf --s subj --trgsubject fsaverage --hemi lh --sval thickness --tval subj_thickness_fsaverage.mgh b) I averaged the thickness or area data within the ROI in fsaverage space for each subject using mri _segstats mri_segstats --seg ROI_lh.mgh --in subj_area_fsaverage.mgh --sum output_area_subj.stats mri_segstats --seg ROI_lh.mgh --in subj_area_fsaverage.mgh --sum output_thickness_subj.stats c) I used a 2-sample T test to compare the area or thickness output between my two groups of subjects. Method 2: a) I transformed ROI.mgh into a label file ROI.label mri_cor2label --i ROI_lh.mgh --id 1 --l ROI_lh.label --surf fsaverage lh b) I transformed the label file from fsaverage space to each subject's space mri_label2label --srcsubject fsaverage --srclabel ROI_lh.label --trgsubject subj --trglabel ROI_lh_subj.label --hemi lh --regmethod surface c) I averaged the thickness and area data within the ROI in each subject's space using mris_anatomicalstats mris_anatomical_stats -l ROI_lh_subj.label -f output_subj.stats subj lh d) I used a 2 sample T test to compare the area or thickness data (from the output_subj.stats files) between my two groups of subjects. For cortical thickness, i obtain very similar results (not identical, but similar to a hundredth of a mm), but for area, I get completely different results, sometimes opposite direction of effect between the two groups. I have 10 ROIs, and across ROIs, group differences are much more significant in the first than the second method. The areal results are in different units, which is expected (in the first method, they are in mm2/vertex and in the second they are in mm2 because they express the total area of the ROI). But despite that, shouldn't I get similar patterns of results ? I would have assumed that the vertexwise areal values averaged in the first method should somehow reflect differences in area betwen subjects and therefore should give similar result to the total area approach (2nd method). Or is it that, in the first method, mri_surf2surf interpolates the vertexwise areal values without taking into account the stretching/deformation from the subject's space to fsaverage space (in contrast with a vertexwise analysis using mris_preproc) ? This is suggested by the fact that when I divide the total ROI area data obtained in the second method by the number of vertices in the ROI (in each subject's space), I end up with results that are very similar to the first method. The absolute values are different (around 0.7 mm2/vertex in the first method and around 0.6 mm2/vertex for the second method), but the group effects are in the same direction (across the 10 ROIs, all the effect sizes and p-values are similar so it can't be a coincidence). This suggests that the vertexwise areal values obtained in both methods simply reflect the ROI-average vertexwise areal values of each subject's mesh (and therefore do not include potential areal differences between subjects, assuming similar vertex spacing betwen subjects, and so shouldn't be used to compare areas between groups). However, if this is correct, isn't it strange that I should find significant differences in these vertexwise areal values between groups? So I guess I have three main questions: 1) Why is there a difference between the 2 methods for area but not thickness ? 2) What is the 'correct' method for area (if any) ? Or is there a better method for an ROI analysis of area? 3) Why are there even significant differences between vertexwise area values in the first and second method ? I am using Freesurfer v 5.3.0 (although the first analysis was potentially run using 5.0.0, not sure!) Thanks in advance Julien -- ------------------------------------- Julien Besle Senior Investigator Scientist MRC Institute of Hearing Research University Park Nottingham NG7 2RD United Kingdom tel: +44 (0)115 951 8508 Ext 209 ------------------------------------- _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.