FreeSurfer Tutorial and Workshop Quiz
Discuss the answers to these questions with your partner.
What is the difference between a volume and a surface?
- A volume stores information about 3D space, whereas a surface stores 2D space.
Do I have to use two MPRAGEs to run recon-all? Explain.
- No, you do not have to use two structural scans as input to recon-all. If a multi-channel head coil is being used, then generally the SNR will be high enough where averaging two scans is no longer beneficial (and can even worsen a perfectly good scan).
Can FreeSurfer help me select a region of interest and measure certain quantities within? How?
- You can draw a region of interest either in the volume (using tkmedit and the Select Voxels tool) or on the surface (using tkmedit or qdec).
What is fsaverage made of?
- It is constructed of MRI scans of 40 subjects that have been manually segmented. 10 are young adults, 10 are middle-ages, 10 are older cognitively normal, and 10 are older demented patients.
What measures will FreeSurfer give me?
- Volume, Mean Intensity (plus standard deviation, min, and max) of the Subcortical structures, Cortical gray matter, CSF, Cerebellum, White Matter, Ventricles, and Brainstem. Number of vertices, Surface Area, Volume, Thickness, Mean Curvature of the parcellated cortex. Also, Total Gray Matter Volume (cortical + subcortical gray), Supratentorial Volume (everything above the tentorium), Total Cortical Gray Volume (difference between pial and white surface), Subcortical Gray volume (total of all subcortical gm volumes), Total White Matter volume (volume within white surface excluding subcortical gray matter), estimated ICV.
How long does it take for recon-all to finish processing one subject?
- 20-24 hours.
How long would it take you to do this manually?
- To manually label the wm, gm, and subcortical structures of one case will likely take a month.
Why do I have to set so many variables before using FreeSurfer?
- The variables indicate the location of code and data sets. Setting them once at the beginning of your data processing ensures that the right version of the code is executed, all the binaries will be found and that the data files are also correctly located.
Where do I find all those fantastic stats files FreeSurfer created for me?
The stats output for each subject can be found in <subjid>/stats where subjid is the the name of one particular subject. To grab the stats files for several subjects to put them in spreadsheet-ready format, use the commands asegstats2table or aparcstats2table.
When mailing the FreeSurfer list about a problem, what information should I include?
You should include the version of FreeSurfer you are using, the command line you tried to run, the error message you got either in the shell window or the recon-all.log, and the Operating System you are running FreeSurfer on. You can find out which version you are running by typing:
Why is spherical averaging better than current volume-based methods out there?
- Current volume-based registration methods cannot achieve high accuracy in aligning the cortical regions because of their high inter-subject variability. When we are only interested in finding correspondences between these areas (for example, in the case of functional studies), it is sufficient and also more accurate to register only the cortical areas with the spherical averaging method.
What is a limitation of this procedure?
- The alignment does not take into account the subcortical structures.
When you do all that crazy morph stuff, are you changing the data?
- When computing the deformation fields between a subject and another or an atlas, the data itself is not changed. When the deformation field is computed, however, it can be applied to the subject, which means resampling its scan(s) in the target coordinate space.
Do I have to remove every bit of skull, dura, etc. that I see in the brainmask? Why or why not?
- You only need to remove skull or dura if it affects the surfaces since we get our measurements from the surfaces.
Oh no! I made all these edits to a subject but now I want to rerun recon-all on the subject again with a new version of FreeSurfer. Will I lose all my work?
If you have already made edits to a dataset, you can rerun it with a new version of FreeSurfer and it will keep all of the edits you have already made as long as you run it on the existing dataset. In other words, you would just want to do "recon-all -all -s existing_subject".
Where do all those atlases come from that FreeSurfer uses?
The different FreeSurfer atlases were generated from manually segmented data sets. Given each definition of the manual segmentation procedure and its labels, we constructed an atlas.
If a ran my subjects with version 4.0, can I run the rest with the newest version?
For population studies and large scale comparisons, the best is to process all your data sets with the same version of FreeSurfer.