Inspection of Freesurfer Output
In this exercise you will visualize and inspect correctly processed output data so that you can become familiar with what the end product should look like. The exercise will take you through a variety of output, but is not necessarily the recommended procedure to take when trying to verify each subject for a real study. Some outputs are only necessary to check when troubleshooting, for example. However, it is a good idea for new users to become familiar with the variety of outputs and how to view them.
If You're at an Organized Course
If you are taking one of the formally organized courses, everything has been set up for you on the provided laptop. The only thing you will need to do is run the following commands in every new terminal window (aka shell) you open throughout this tutorial. Copy and paste the commands below to get started:
setenv SUBJECTS_DIR $TUTORIAL_DATA/buckner_data/tutorial_subjs cd $SUBJECTS_DIR
To copy: Highlight the command in the box above, right click and select copy (or use keyboard shortcut Ctrl+c), then use the middle button of your mouse to click inside the terminal window (this will paste the command). Press enter to run the command. These two commands set the SUBJECTS_DIR variable to the directory where the data is stored and then navigates into this directory. You can now skip ahead to the tutorial (below the gray line).
If You're not at an Organized Course
If you are NOT taking one of the formally organized courses, then to follow this exercise exactly be sure you've downloaded the tutorial data set before you begin. If you choose not to download the data set you can follow these instructions on your own data, but you will have to substitute your own data paths and subject names. These are the commands that you need to run before getting started:
tcsh source your_freesurfer_dir/SetUpFreeSurfer.csh setenv SUBJECTS_DIR $TUTORIAL_DATA/buckner_data/tutorial_subjs cd $SUBJECTS_DIR
Notice the command to open tcsh. If you are already running the tcsh command shell, then the 'tcsh' command is not necessary. If you are not using the tutorial data you should set your SUBJECTS_DIR to the directory in which the recon(s) of the subject(s) you will use for this tutorial are located.
Viewing Volumes with Freeview
With one Freeview command line, you can load several output volumes, such as brainmask.mgz and wm.mgz; the surfaces, rh.white and lh.white; and the subcortical segmentation, aseg.mgz. Copy and paste the command below inside the terminal window and press enter:
freeview -v \ good_output/mri/T1.mgz \ good_output/mri/wm.mgz \ good_output/mri/brainmask.mgz \ good_output/mri/aseg.mgz:colormap=lut:opacity=0.2 \ -f good_output/surf/lh.white:edgecolor=blue \ good_output/surf/lh.pial:edgecolor=red \ good_output/surf/rh.white:edgecolor=blue \ good_output/surf/rh.pial:edgecolor=red
Some notes on the above command line:
- good_output is the name of the subject
- The flag -v is used to open volumes
- brainmask.mgz : skull-stripped volume primarily used for troubleshooting
- wm.mgz : white matter mask also used for troubleshooting
- aseg.mgz : subcortical segmentation loaded with its corresponding color table and at a low opacity
- The flag -f is used to load surfaces
white & pial surfaces are loaded for each hemisphere & with color indicated by 'edgecolor'
After hitting enter, a freeview window should open showing you the outputs you specified:
The menu on the left shows which files have been loaded.
Use the buttons at the top to change which orthogonal view appears in the main viewing window. Let's go with coronal for now. You can also use the buttons to change the organization of the viewing panes. To change which brain slice you are viewing, use the 'Page Up' or 'Page Down' keys on your keyboard. (Mac users: press the fn key while using the up and down arrows.)
While Freeview can load many volumes at once, you cannot necessarily see them all at once. You are able to see whichever volume is at the top of the list in the menu on the left. An exception to this are volumes such as the wm.mgz and aseg.mz which can be made translucent, allowing you to view the information they contain simultaneously with the volume directly below it on the list. For example, you are currently seeing information from both the aseg (labeled structures) and the brainmask (voxel intensities).
You can hide or turn off a layer by unchecking the check box next to the layer name. Try this out - uncheck the box next to 'aseg'. Now you can see just the brainmask. You can also use the up and down arrows (located below the menu on the left) to move the aseg down on the list, below the brainmask (try it!). Let's now move the wm volume to the top of the list but instead of using the arrows, try this shortcut: double click on where it says 'wm'. It should automatically move to the top. The menu should now look like this:
and the viewing window should look like this:
Keyboard Shortcut: Alt-c will allow you to quickly cycle through all the layers. Every time you hit it, the volume at the top of the list will move to the bottom of the list.
When the Navigation button is chosen, you can move the image in the viewing window around by holding down the middle mouse button and dragging the mouse where you want the image to go. Try it out. You can also move the image more slowly using the up/down and left/right keys. To zoom, scroll with the middle mouse button. In this navigation mode, notice the cursor (little red crosshair) moves to wherever you left click. When you change the orientation (to axial or sagittal), you will be viewing the slice that intersects with the cursor's location. To illustrate this point, if you are in the viewing pane selected here:
You'll notice all the planes will shift based on where you move the cursor.
One other important thing to note is that any action you do in the viewing window (i.e. erasing, changing brightness, etc.) will take place on whichever volume is currently highlighted in the left menu, regardless of which file is at the top of the list.
Now that you know the basics, you can make your way through the data. To verify that FreeSurfer did a good job, you will want to check:
- Whether the surfaces accurately follow the gray matter and white matter boundaries.
- Whether the aseg accurately follows the subcortical intensity boundaries.
Checking the Surfaces
Double click on 'brainmask' in the left menu to bring it to the top of the volume list. The white surface (blue line) is used to calculate total white matter volume and should accurately follow the boundary between white matter and gray matter. The pial surface is used to calculate cortical gray matter volume and should accurately follow the boundary between the gray matter and the CSF.
As you scroll through the slices checking the surfaces for accuracy, keep in mind that you are looking at a 2-dimensional rendering of a 3-dimensional image - be sure to look at more than just one view (i.e., sagittal, coronal and horizontal). You can turn the surfaces off and on by checking and unchecking them in the left menu under where it says 'Surfaces'. As you do this, ask yourself: would you draw the boundary in the same location?
Keyboard Shortcut: Alt-f will turn on and off whichever surface is highlighted in the menu window.
To help verify accuracy, adjust the brightness and contrast so you can easily identify the shift in intensity between gray and white matter. To do this, left click on the image while holding down the 'Shift' key and drag your mouse. (Make sure the brainmask volume is highlighted in the left menu in order for this to work.) The other way to do this is via the 'Window' and 'Level' sliders underneath the left menu.
There are regions where the surfaces are not intended to be accurate that you should be aware of:
- Areas around the hippocampus and amygdala. The surfaces will not completely include or exclude certain subcortical regions. These inaccuracies can be ignored as subcortical regions are excluded from the cortical measures and subcortical volume is measured by the aseg, not the surfaces.
- For an example of this, scroll to coronal slice 137 (slice numbers appear in the upper right hand corner of the viewing window).
- Along the midline cut, it is possible to see some overlapping of the surfaces from one hemisphere to another. The medial wall is not included in the cortical measures so this can generally be ignored.
Double click on the aseg so it is directly above the brainmask in the left menu.
This will show the complete segmentation of the subcortical structures.
Each structure is labeled with a unique color/number distinction. If you click on a voxel the structure's name and number label will be shown in the 'Cursor' section under the viewing window next to the word, 'aseg'. If you hover over a voxel where the cursor is not located, the value of that voxel will appear under the 'Mouse' section.
Keyboard Shortcut: Alt-v will turn on and off the layer that is currently highlighted.
Make sure 'aseg' is highlighted in the left menu and press Alt-v to turn it off and on. While doing this, make sure the aseg is accurately following the underlying intensity boundaries of each structure. You can also adjust the 'Opacity' slider to better see the underlying brainmask.
Keyboard Shortcut: Alt-a and Alt-s will change the opacity of the layer that is currently highlighted.
Other Things To Do in Freeview
Below, we introduce you to other volumes and techniques which are good to be familiar with, especially when troubleshooting.
First, close the aseg volume by highlighting it and the clicking this button . Then, to load the aparc+aseg segmentation go to File > Load Volume, click on the yellow folder icon, and browse to the aparc+aseg.mgz. Once you have selected the file, click 'Open', and then 'Ok'. When it loads, 'Grayscale' will be selected as the Colormap. This will obscure the brainmask beneath. In the Color map section on the left, choose 'Lookup Table'. Then adjust the opacity so that you can see the brainmask and aparc+aseg at the same time. When loaded, it will look like this:
This volume shows the parcellated cortical ribbon at the same time as the segmented subcortical structures (later we will look at the cortical labels on the surface). Click around the cortex to see the name of each cortical region under the 'Cursor' section. The aparc+aseg.mgz uses the Desikan-Killiany atlas. To see the Destrieux atlas, you would load the aparc.a2009s+aseg.mgz
Close the aparc+aseg. As you scroll through the brainmask volume, notice that there is no skull left in your image. You should also not see any regions of cortex or cerebellum missing from this volume. Bring the T1.mgz to the top of the volume list and toggle between it and the brainmask.mgz volume (Alt-v) to verify that the skullstrip has worked properly. In the TroubleshootingData tutorial, we'll go over what to do if there was a skull strip error.
Scroll through the brainmask volume and notice that the intensity is all uniform. You should not see any very bright or very dark spots in the white matter or gray matter. If you click on a voxel in the white matter, you can see that it has been normalized to an intensity of (or very close to) 110 (look under the Cursor section next to where it says 'brainmask'). When wm voxels are far from a value of 110, they may be erroneously excluded from the white surface. In the TroubleshootingData tutorial, we'll go over what to do if there was an intensity normalization error.
Double click on the wm volume to bring it to the top of the list. This volume is FreeSurfer's initial segmentation of the white matter (shown in gray) with additions from the automatic topology fixer (in white). This "mask" is the starting point for the white surface which grows out from here and stops at a more accurate location using the intensity gradients in the brainmask.mgz volume as a guide. The wm.mgz can be used to add missing wm voxels or delete voxels that are not white matter but were included in the surface. In the TroubleshootingData tutorial, we'll go over how to do this and when. For now, it will be good to learn how to change the wm mask to a heat overlay for ease of editing. With wm highlighted in the left menu, take a look at the options next to Color Map and choose 'Heat' or 'Jet'. Then adjust the opacity so you can also see the brainmask underneath (down to around .25). It should look like this:
This could have also be done via commandline when first loading the wm in Freeview if we used this command: freeview -v wm.mgz:colormap=heat:opacity=0.25 brainmask.mgz (Note: You don't need to run this command.)
You can now close Freeview by hitting the X on the display window or Ctrl-q.
Viewing Surfaces in 3D using Freeview
We're now going to view several surface overlays. You could view the volumes discussed above and the overlays discussed below all in one Freeview session. They are separated in this tutorial only for simplicity. The examples below are displayed only on the left hemisphere, however, you could also view just the right hemisphere or both hemispheres at the same time. Here are some surfaces you can look at with Freeview:
- pial, white and inflated surface
- sulcal and curvature maps
- thickness maps
- cortical parcellation
You can load them all in Freeview with the command below (be patient while they all load):
freeview -f good_output/surf/lh.pial:annot=aparc.annot:name=pial_aparc:visible=0 \ good_output/surf/lh.inflated:overlay=lh.thickness:overlay_threshold=0.1,3::name=inflated_thickness:visible=0 \ good_output/surf/lh.inflated:visible=0 \ good_output/surf/lh.white:visible=0 \ good_output/surf/lh.pial
Some notes on the above command line:
lh.pial:annot=aparc.annot loads the Desikan-Killiany cortical parcellation on the pial surface
:name=pial_aparc:visible=0 changes which name shows up in the menu display and turns off this layer
lh.inflated:overlay=lh.thickness:overlay_threshold=0.1,3 loads the thickness overlay on top of the inflated surface and sets the min and max thresholds to display
Pial Surface The first volume you see is the pial surface. The pial surface here is the full 3D representation of the red surface you saw on each 2D slice of the brainmask volume, earlier in the tutorial. The green regions are gyri and the red regions are sulci. With this surface, the sulci are mostly hidden. Feel free to move the inflated brain around by left clicking on it and dragging the mouse. If you would like to put the brain back to its original state, go to View > Reset View or press Ctrl+r.
White Surface Press Alt+c to cycle to the white surface.
The white surface shows the boundary between white matter and gray matter. It is the 3D representation of the blue surface you saw on the 2D slices of the brainmask. With this surface, we are able to see the sulci a bit better. You can inspect this surface by rotating it around as you wish.
Inflated Surface Press Alt+c to cycle to the curvature on the inflated surface. With the inflated surface, you can fully see the sulci. If you look at the options next to 'Curvature' in the left menu pane, you can switch it to binary to see the curvature in grayscale. Or switch it to 'off' to see the smooth inflated surface.
The inflated surface can be helpful to find bumps, holes, or other defects that may need to be corrected. If you click on the inflated surface, you will see the coordinates of the vertex you clicked on in the 'Cursor' window.
Thickness Map Press Alt+c to cycle to the thickness map on the inflated surface.
Cortical Parcellation Press Alt+c to cycle to the cortical parcellation.
Click on a color and view the name of the cortical region in the 'Cursor' or 'Mouse' windows. Rotate the brain so you are looking at the medial wall. Notice that all subcortical gray matter is not a part of the surface labels (because again, those areas do not count towards the cortical surface measures). The parcellation that is loaded here was created with the Desikan-Killiany atlas. By default there are two parcellations that are made when recon-all is run. The second parcellation, called ?h.aparc.a2009s.annot, is created with the Destrieux atlas. The difference is the number and designation of the areas that are labeled. You can load this second parcellation by first going to File -> Label -> Delete all Labels. This will remove the first parcellation. Then you can repeat the steps for loading a parcellation, this time browsing to lh.aparc.a2009s.annot to load the second parcellation.