Dear Xin-Fa, > 1. I find too much files in JHU and Juelich atlases folders of FSL, and > which one is of help ? > You may want to open the .xml files to see which structure corresponds to each index, then locate them in the "prob" files (these are 4D files). > 2. How to got the masks I need ?Could FSL have some masks or make masks? > How to make mask ? > You'll need tools like fslsplit and fslmaths to get them as separate images. The FSL commands are pretty straightforward and you may want to do the tutorial to become familiar with them all. > 3. In fact I just wonder how to make further study of significant > regions resuting from TBSS ,for example their name,coordinates etal as > shown Fig.1. > Once you have the mask(s), the values can be extracted from the original data with the command fslmeants. > 5. Correlation analysis between FA and covariants is conducted by GLM > and randomise, is right ? If so ,how to got essentially the same > information as correlation coefficient r and P ? > A correlation tells how well one set of data explains another, and so does the t-statistic. Their p-values have equivalent interpretation. You can just randomise and report the p-values you find with it (which indeed is the recommendation). That said, however, it seems that your Fig 3 is the scatter of some variable vs FA measured either in a voxel or the average over an ROI. You can still use randomise, but note that the input needs to be in NIFTI format, even if it's not really a full image. Perhaps it's simpler if you just compute the correlations using some spreadsheet software, like LibreOffice or Excel. All the best, Anderson 2015-01-08 17:59 GMT+08:00 Anderson M. Winkler <[log in to unmask]>: > Dear Xin-Fa, > > Please, see below: > > On 7 January 2015 at 14:05, SUBSCRIBE FSL Xinfa_Shi < > [log in to unmask]> wrote: > >> Hello all, >> >> I have finished TBSS pipeline according to user guide ,and got some >> significant regions .However , I still have the following questions to ask >> for help . >> >> Before that I am sorry about my poor English . >> >> 1. Which atlases of FSL correspond to JHU-ICBM-DTI-81 White Matter and >> Juelich Histological atlases ,which were used to label clusters with >> significant FA alterations . >> > > These are the "JHU ICBM-DTI-81 White-Matter Labels" and "Juelich > Histological Atlas" respectively. > > >> 2. How to extract the FA value of every subject and average FA value of >> per group , and then plot something similar to figure 1? >> > > You need a mask, either for the whole brain or for some region of > interest, then use the command "fslmeants". > > >> >> 3. How significant difference regions can be made into ROIs , and these >> ROI can be applied to every subjects ? or say how to implement a ROI-based >> TBSS? >> > > An ROI can be made after thresholding a map with fslmaths, but using it to > do further stats can be circular and invalid. Consider instead an ROI > defined independently from the data, e.g., from an atlas. > > >> >> 4. Could TBSS make analysis with some covariants(e.g. age,weight)? If so >> ,please tell me way because I just make simple analysis whitout any >> covariant following user guide . >> > > Yes, there are examples in the GLM manual, here: > http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM > > >> >> 5. Could TBSS conduct correlation analysis between FA and >> covariants(e.g. age,weight) ,and get correlation coefficient r and P ,as >> shown in figure 3 ? >> > > Yes, but in general r isn't produced, but a t-statistic instead, which > provides essentially the same information. > > All the best, > > Anderson > > > > > >