• The evolution of the arcuate fasiculus revealed with comparative DTI - Rilling et al., 2008
• Subcortical connections to human amygdala and changes following destruction fo the visual cortex - Tamietto et al., 2012

• Observe variability in brain size and shape between individuals
• Transform individual subject brains to MNI space with FSL Flirt and compare to a brain atlas.
• Average individual subject transformed brains using Matlab.
• Learn about statistical brain atlases, Talairach, and MNI (Montreal Neurological Institute) space.
• Create anatomical region-of-interests with the FSL atlases.
• Superimpose the ROI on the brains you transformed into MNI space.
• Use FSL FIRST to segment and label subcortical brain regions such as hippocampus and amygdala
• Examine a brain you labeled with FreeSurfer

• AFNI image viewer
• FSL's Flirt (FSL Linear Registration Tool) for image registration.
• Learn to write a simple AFNI script in bash to average brains.
• FSLeyes atlases
• FSL/FIRST to label subcortical structures
• FreeSurfer for individualized atlases.

• none

Many scientists and clinicians would like to compare the brains of different individuals and make comparisons and averages of quantitative measurements of different brain structures. For example, a clinical psychologist might be interested in whether a brain structure such as the hippocampus is different in size for depressed compared to non-depressed individuals. A genetics researcher might want to know if different alleles of the serotonin transporter gene are associated with different sized amygdala. A language researcher might want to know if there are hemispheric differences in temporal lobe anatomy in individuals who have right or left language dominance. These examples raise an important question - How different are the brains of individuals?

In the first part of this exercise, you will closely observe three brain regions in three or more brains that have been previously skull-stripped. You can select three of the five brains available in ~/Desktop/input/mri/anat_highres.

1. Open your Terminal app and change your directory to /Users/hnl/Desktop/input/mri/anat_highres

2. Open FSL

3. Open three instances of FSLeyes.

• Each time you click the FSLeyes button, a new instance will open.

4. Open a different skull-stripped brain in each instance. (Click here if you forgot how to load brains in FSLeyes.)

5. Get your windows nicely organized and sized appropriately so each brain looks about the same.

Let's spend a few minutes familiarizing ourselves with a few features of FSLeyes.

1. You can learn a bit more about the brains you've loaded by clicking on the info button. A window will open with information, including the dimensions of the image (and therefore, voxel size).

2. You can change the arrangement of your three brain views (axial, coronal, and sagittal) within the window with these three buttons:

3. You can turn on/off your different brain views (axial, coronal, and sagittal) with these three buttons:

The next tips will be particularly helpful for you when writing lab reports!

4. By default, you see green crosshairs overalid on top of your brain indicating your current specific voxel location in the three views. But sometimes it's better to see the brains without these crosshairs. Use the button below to toggle then on/off.

5. You previously learned how to take screenshots using the macos. But FSLeyes has a builtin screenshot function that you'll find helpful when you don't want all the “extra” stuff (e.g., menu bars) in your screenshot.

6. In one of the windows, click on a precise anatomical location of your choosing.

• Copy the X,Y,Z voxel coordinates of that location (see the red boxes in the image below) into the 2D viewer of the other two brains.
• When you are done, your crosshairs will be at the identical matrix position in all three windows of FSLview.
• Are you at the same anatomical location when you are at the same coordinate?

7. To help focus your comparison, find a (relatively) similar slice. For example, in the image below I centered my crosshairs in the midline of the anterior commissure. You can use this, or any other, anatomical landmark to help you get similar slices in each of your three brains.

• Spend a few moments exploring the the overall similarities and differences among the brains.

To automate measuring brain structures and differences in brain structures, it would be very helpful to have those structures at the same coordinates. Although not our problem for today, putting brain regions into the same coordinate system is essential for most functional MRI group statistical analyses. So, we are now going to do this.

We will transform at least two of the skull-stripped brains from the last exercise to a common coordinate system (the MNI coordinate system, named for the Montreal Neurological Institute). We will use the FSL module flirt (FSL Linear Registration Tool) to accomplish this. Flirt will create transformation matrices that can then be used to align the brains to the common space. We will then quickly repeat Part 1 comparing the two individual brains after registration.

1. If you closed FSL after the last exercise, reopen it by typing fsl & in your Terminal window.

2. Click on the FLIRT linear registration button.

3. Set the reference image to MNI152_T1_1mm_brain. By default it is set to MNI152_T1_2mm_brain, so you can just change the 2 to a 1.

4. For the Input image, specify one of your skull stripped brains from Part 1 (i.e., a brain from the ~/Desktop/input/mri/anat_highres collection)

5. Specify the output. Start with /Users/hnl/Desktop/output/lab05/ (this is the folder in which your output will go). You will also need to add your output file name to the end of this path (see tip box below).

6. Change the Model/DOF (input to ref) option from Affine (12 parameter model) to Rigid Body (6 parameter model).

7. Press Go when you are done.

8. Once this brain is complete, carry out the same operation on a second individual's brain, again using a 6 DOF model. Make sure to give this second one a unique output file name otherwise you will overwrite the first one that you ran.

We are now going to see how well the transformations worked.

In this section we will use yet another image viewer; AFNI. AFNI is my preferred software package for fMRI analysis, but it is ugly and relatively difficult to learn. However, the viewer has a nice feature that will allow us to “lock” the view of different brains.

1. Open your Terminal app and change your directory to your output directory ~/Desktop/output/lab04.

2. Copy your reference image into your current directory using the following command in your Terminal window (we have to do this because AFNI doesn't work well when using images located in different directories).

cp /Users/hnl/Desktop/input/mri/fsl_standard/MNI152_T1_1mm_brain.nii.gz .

2. Start AFNI by typing afni &

• By default AFNI will
  • load the first available brain in the directory
  • open the AFNI control window
  • open two views of the brain in two separate windows; axial and sagittal

3. Change the brain to one of your transformed brains.

• click Underlay
• choose the file from the dropdown menu
• click Set

4. Open a new viewer and load your second transformed brain.

• to open a new viewer, click on the New button in the bottom left of the control window
• in the new control window that opens, load a different skull-stripped brain (see #3 above)
• the new control window does not automatically open any views of the brain. So you'll need to click on the Image button next to Axial and Sagittal

5. Open a third viewer and load the MNI152_T1_brain.

6. Get your windows nicely organized and sized appropriately so each brain looks about the same. Unlike FSLeyes, AFNI opens each brain view in a separate window. This can make window management a bit of a pain. But you should try to get your windows organized so that they look something like the image below. Note that the title of each window starts with [A], [B], and [C]. These correspond to the first, second, and third control windows you opened.

7. To evaluate how well the transformation worked, click on three or more structures in the MNI reference brain and see if the cross-hairs show up in those same structures in your 'flirted' brains.

• Try the amygdala (one hemisphere only).
• Try the head of the caudate.
• Try the calcarine sulcus in the occipital lobe.

1. Now that you are familiar with 6 DOF models, try experimenting with a different DOF model run on a single subject. Remember to have a different output filename for this output. I suggest appending _reg12.

• In Part 3 of this lab you ran the 6 DOF model. Now FLIRT the same subject using the 12 DOF model.
• Make sure to have your output go into the same directories you used for the 6 DOF, so that AFNI will have access to them.

2. Open the files that were generated with different DOFs in 'locked' AFNI windows.

• The four AFNI viewers should display
  1. the MNI152 template brain
  2. your 6 DOF FLIRTed brain
  3. your 12 DOF FLIRTed brain

FSLeyes provides nice set of probabilistic atlases that you can overlay on the standard brain to understand the name given to different anatomical locations.

1. Open the MNI152 brain in FSLeyes

• Start FSLeyes
• Select File → Add Standard
• Choose MNI152_T1_1mm_brain.nii.gz

2. Click on Settings → Ortho View 1 → Atlas Panel

3. Now if you click around to different areas of the brain, information will be updated in the atlases tools box that tells you about the anatomical location at the cross-hair. For instance, in the crosshairs in the image below are on the Right Hippocampus according to the Harvard-Oxford Subcortical Structural Atlas.

You can add or remove atlases from the list.

1. To add or remove atlases, simply click on the checkbox to the left of the atlas name.

2. I recommend adding the 'Juelich Histological Atlas' and 'Talairach Daemon Labels' to start out (keep the current Harvard-Oxford atlases). It will take a few seconds for FSLeyes to update atlas tools with the new labels.

3. If you want to find out more about the different atlases, you can go to this site.

You can visualize specific brain regions or create anatomical regions of interest (ROI) by:

1. Clicking on the 'Atlas Search' tab inside of the 'Atlases' window.

2. You can select your atlas by clicking on the checkbox to its left. Select 'Harvard-Oxford Cortical Structural Atlas'. You will see a colorful overly appear on your brain. Each color represents a different cortical region according to the atlas.

3. You can now jump to a specific structure. To find a structure you're interested in you can either scroll through the (long) list, or simply start typing a name into the Search box. Below I entered 'fusiform' in the Search box.

4. If you click the + sign to the left of the region name, you FSLeyes will center your cross hair to where the region resides in the standard brain. Very helpful! Try it now by clicking on 'Temporal Occipital Fusiform Cortex'. You should have jumped to that region as in the image below.

5. You can also overlay the probabilistic values for this region from the atlas. Click on the checkbox to the left of the + sign (to the left of 'Temporal Occipital Fusiform Cortex'). Do so now.

• Your display will look like the one below. Each voxel is now shown with an associated probability that it is in the selected region. The more red the voxel, the more likely that it's in the 'Temporal Occipital Fusiform Cortex'.

6. Notice the Min and Max values at the top of the window. These values determine the threshold for what is displayed. The values have different meanings depending on what type of brain/image you're looking at.

• If you highlight MNI152_T1_1mm in the Overlay list you'll see that values default to 0 and 8447.64, which are arbitrary intensity values associated with the greyscale brain.
• If you highlight harvardoxford-cortical/prob/Temporal Occipital Fusiform Cortex, the values default to 0 and 95.95. This means you are displaying all voxels that have a probability between 0-95.95% of being in the region (the default max value tells us that there are no voxels with a higher probability than 95.95% in the brain).
  • Change the Min value to something higher, like 50. You'll see that you're region gets smaller, because now you're being more conservative and restricting the display to only show (i.e., color) voxels that have a 50% or higher probability of being in the region.

The FSL Atlases overlaid a group derived anatomical brain mask upon your transformed brain. Is this as accurate as determining the brain structure from an individual, non-normalized brain? In other words, is it as accurate to say “we think this region is Prof. Engell's amygdala because it's the amygdala in 94.5% of individuals” as saying “we think this region is Prof. Engell's amygdala because we identified it in his brain image”? We can investigate this question by using FSL ''FIRST'' - a program that automatically segments and labels sub-cortical structures.

The FSL FIRST program can only be run from the command line in your Terminal app. You can specify your skull-stripped brain as input, or you can choose any of our sample brains. Note, that you can specify a brain that has NOT been skull stripped. In this case, the FIRST program will do the skull-stripping by calling the BET program. However, since we like to make sure our skull-stripping is of high quality, we will specify a brain for which we have supervised the skull stripping ourselves. However, you need to tell FIRST that the brain is skull stripped - you do that by specifying the -b option in the command line.

In the code below, I used the skull-stripped version of the 34353 brain located in the ~/Desktop/input/mri/anat_highres directory.

1. In this example, we will create short-cuts in bash to point to the input and output directories. Be careful not to put any spaces before or after the = or the space will become part of the name. By convention, we use capitals for such short-cuts, but it is not required.

INPUTDIR=~/Desktop/input/mri/anat_highres
OUTPUTDIR=~/Desktop/output/lab04

We can then use these short-cuts in our command line. They will be “expanded” when the command line is executed to the paths to which we equated them.

echo $INPUTDIR

2. Notice, to let the bash shell know that we are specifying short-cuts, we prefix the short-cut with a $ when we use it.

run_first_all -v -b -i ${INPUTDIR}/34353_highres_skullstripped.nii.gz -o ${OUTPUTDIR}/34353_

The command line above executes the run_first_all FSL script. This script calls several programs that comprise FIRST. Note that there are four options on the command line:

• -v for verbose output
• -b to indicate that we are submitting a skull-stripped brain as input
• -i specifies that the next item is the input file
• -o specifies that the next item is the path and prefix of the output files. There will be several output files, and each will start with the specified prefix '34353_'. You can replace that prefix with anything you want.

3. Here is what you should do with the output:

• Overlay the segmented output (e.g., 34353_all_fast_firstseg.nii.gz) from FIRST upon your original skull stripped brain using FSLeyes.

So how did the MNI152 brain come about anyway? Somebody at the Montreal Neurological Institute co-registered 152 brains to a single brain, and then averaged across all the registered brain. Let's try this on the five brains for which we have good registrations, and we will create the Kenyon5 brain!

1. We will create an average brain by running the bash script below. This script will call the AFNI command 3dcalc.

3dcalc \
-prefix /Users/hnl/Desktop/output/lab04/kenyon5_brain.nii.gz \
-a /Users/hnl/Desktop/input/mri/anat_highres_trans/34353_12dof.nii.gz \
-b /Users/hnl/Desktop/input/mri/anat_highres_trans/34433_12dof.nii.gz \
-c /Users/hnl/Desktop/input/mri/anat_highres_trans/34532_12dof.nii.gz \
-d /Users/hnl/Desktop/input/mri/anat_highres_trans/34554_12dof.nii.gz \
-e /Users/hnl/Desktop/input/mri/anat_highres_trans/34569_12dof.nii.gz \
-expr '(a+b+c+d+e)/5'

• The \ backslash at the end of each line tell bash that the command continues on the next line
• -prefix tells 3dcalc what we want to name the output file
• Each of the file names is assigned to a unique letter from a to e
• -expr tells 3dcalc what we want done with the input files. In this case we calcualte a simple mean of the five datasets.

3. Open the Kenyon5 brain and the MNI152 brain in locked AFNI windows. Refer back to Part 3 if you need a refresher on how to do this.

FreeSurfer is a powerful segmentation and automatic labeling program that has some similarities to FSL FIRST. However, unlike FIRST, FreeSurfer labels the entire brain, and does quite a lot of other things, such as creating inflated brain surfaces. FreeSurfer can take ~30 hours to run one brain, so it doesn't make for a good in-lab exercise. However, exploring the output can be fun, as you will see below.

• freesurfer needs environment variables set to point to the data. This must be done in the terminal.

1. Type the following codes to a terminal window.

  # set the location of your subject's directory
  export SUBJECTS_DIR=~/Desktop/input/mri/freesurfer_test
 
  #
  export doublebufferflag=1
 
  # set the subject ID
  export SUBJID=subj19_1A
 
  # Tell the viewer to display the pial view of the left hemisphere
  tksurferfv $SUBJID lh pial

The following image should pop up.

You can use the notes below with the accompanying image to get an idea of the different buttons for tksurfer.

Through the TkSurfer Tools GUI window, you can click on different surface views (main, inflated, etc.)

You can view the curvature of the brain via a green-red colormap: Green indicates a gyrus, Red indicates a sulcus.

2. Through the GUI:

• File → Curvature → Load Curvature… item to load a curvature file.
• Choose lh.curv (it is likely the default option so you can click ok)
• You can turn the curvature on/off by clicking on the curvature button on the GUI

Or use the following command from the terminal:

  tksurfer $SUBJID lh pial -curv lh.curv

Next, load the annotation (label) to view the segmentation overlay on the cortical surface:

3. Through the GUI:

• File → Label → Import Annotation…
• Choose one of the .annot files for the hemisphere you are viewing. For example, you can try lh.aparc.annot.
• Click on different areas and the label is displayed on the bottom right corner of the “TkSurfer Tools” window.

For reference, here is a labeled image of a sample freesurfer brain done by the freesurfer folks: