• Spatial Normalization

• You will gain familiarity with image viewers and other software tools we will use throughout the semester.
• You will learn to use software that masks the brain and strips the skull.
• You will understand how the intensity of voxels can be related to different brain tissue types.
• You will examine brain images in 2 dimensional views and create 3 dimensional volume renderings.
• You will search for anatomical brain regions using these software tools.

• MRIcroGL for volume rendering.
• FSLeyes
• FSL and the BET skull-stripping tool.
• FSL and the FAST tissue segmentation tool.

1. Create the output directory for tonight's lab.

mkdir ~/Desktop/output/lab04

• MRI brain data is located in /Users/hnl/Desktop/class/input/mri/mri_retest
• For parts of the lab you will each work on a different brain. When I refer to your “assigned brain” I am referring to those listed below.

There are lots of different MRI viewer programs available. As you might expect they all have their own strengths and weaknesses. We will primarily be using the viewer that is built into the FSL software package; FSLeyes. In this section of the lab I just want you to get familiar with the this viewer and play around with some of the options.

FSL can be executed using a graphical user interface (a GUI) or by typing commands to the terminal. We will use the GUI for most of our exercises, but you should be aware that the command line interface is helpful when writing your own analysis scripts.

The FSL GUI should appear after a few moments. For today's exercises we will use FSLeyes, BET, and the FAST segmentation tool.

1. Click on the FSLeyes button at the bottom of the FSL menu.You should see the following window.

To familiarize yourself with the viewer:

1. Click on File → Add Standard

2. Select the 'MNI152_T1_1mm.nii.gz' image. Feel to choose any other brain image too. Note that 'MNI152_T1_1mm.nii.gz' includes both the brain and skull, whereas 'MNI152_T1_1mm_brain.nii.gz' is only the brain (the skull has been removed from the image).

3. Play around with this standard brain.

Location and Intensity. As you click around, observe the values in the lower-right corner of the screen (see image below).

• In the Location box, each voxel is identified in a three-dimensional space by its x, y, and z position.
• in the window to the right, the intensity of the selected voxel is displayed next to the coordinates.
  • This intensity is reported as an arbitrary unit (i.e., the number itself is meaningless)
  • Sometimes, the value represents a meaningful statistics such as a z-score.

4. Continue to play with different options to see what they do (you won't break anything). See the FSLeyes quick start page to learn what different buttons do.

5. Close this file when you are done having fun and feel comfortable navigating around in FSLeyes (you are having fun…aren't you!?)

1. Click on File → Add from file

2. The file selection window will default to a directory that we don't want. Jump to your desktop by selecting Desktop in the left panel of the window (in the list of “Favorites:). Then click on the class –> input –> mri –> mri_retest directories and select your assigned file.

3. Select your assigned brain from the list and click open

4. In Part 3 of this lab We will remove teh skull (“skull strip”) this brain using the BET program. After we do, we'll want to use FSLeyes to compare our skull-stripped brain to the original brain, so you can minimize the window, but keep your assigned brain loaded in FSLeyes.

As noted above, different viewers have different strengths and weaknesses. In the last part you used FSLeyes to view the MNI standard brain and your assigned brain in 2D. MRIcroGL is particularly good at quickly rendering the brain in 3D. So let's use it to take a quick look at our assigned brain before we continue on to skull-stripping.

1. Open the MRIcroGL program by clicking on the icon in your dock.

.

• You should see something like the image below. (If instead you see a big block of Swiss cheese or a lego brick, don't worry)

2. Open your assigned brain by selecting File –> Open.

• You should see your brain displayed in 2D similar to what we saw with FSLeyes.

3. Let's now view this brain rendered in 3D. To do so, select Display –> Render from the menu bar.

4. Let's play around with MRIcroGL to get a feel for interacting with a brain rendered in 3D.

• If you click on the brain and drag your mouse around you'll be able to rotate the brain.
• You can zoom in and out by dragging your mouse while holding the right-click.
• Try to make Cutout in the brain by manipulating the x, y, and z sliders in the cutout section (you'll need to press Near to get started.)

This 3D rendering is pretty cool. But one thing that you undoubtedly noticed is that the presence of the skull, scalp, and other non-brain tissue obscures the cortical surface of the brain. Indeed, many viewers (including FSLeyes) would render a useless blob if given a brain like this. So in the next section we'll improve things by stripping away all the non-brain stuff.

When rendering a brain volume, the skull, and neck make it difficult to see brain anatomy in 3D. As we will learn, these brain coverings also make it difficult to coregister brains into a common coordinate system. Consequently, removing the skull from the brain is an important first step in many analyses. In this section we will use the FSL Brain Extraction Tool (BET) program to accomplish this.

1. Launch BET by selecting the BET brain extraction button on the FSL menu. You should now see the following window:

2. Your Input image should be the same assigned brain you used in the previous section.

• Click on the folder icon at the end of the Input image line and use the file selection box to navigate to the file. Select the file and press Ok
  • To go up on level in the directory hierarchy, double-click on the .. under “Directories:”

3. You will also need to specify an Output image.

• Click on the folder icon at the end of the Output Image line and navigate to the /Users/hnl/Desktop/class/output/lab04 folder and select Ok.
• In the Output image line add your filename to the end of the path.

There are MANY options for skull stripping. You are encouraged to experiment with the different settings to achieve the best brain extraction. However, you may want to start by choosing the 'Robust brain centre estimation (iterates BET several times)' option on the first drop down menu.

4. Click Go and BET will start stripping. This usually takes a minute or two. You know you are done when the depressed Go button pops out again. Don't be impatient and keep pressing Go, as FSL will dutifully run another instance of the program.

We will now view the skull stripped brain by superimposing it onto the non-stripped version that you already have loaded into FSLeyes from this earlier section of the lab.

1. In the FSLeyes window select File –> Add from file and then navigate to and select your skull stripped brain you just created.

2. Your skull stripped brain is now being displayed stacked on top of your original brain. But it's nearly impossible to tell them apart. To make it easier to distinguish between them, we'll colorize the skull stripped brain.

• In Overlay list, click on your skull stripped brain so that the filename is highlighted.
• Change the colorscale dropdown menu (the menu is pointed out between the arrows in the image below) from Greyscale to Green (or any other color you'd like!)

• Now you can see your original MRI in greyscale, and the skull stripped version of that same MRI overlaid on top and colored Green (or whatever color you chose.)

High quality skull stripping is defined as successfully removing all of the non-brain features (e.g,. skull), while preserving all of the brain tissue. In other words, if you are too aggressive then you will successfully get rid of all the non-brain, but you'll also throw out some brain. If you are too conservative then you will retain all of the brain, but you'll also keep bits of non-brain. The goal is to be a neuroscience Goldilocks; you don't want to remove too much, you don't want to remove too little, you want to remove the amount that's just right.

1. Closely examine your skull stripped brain - does it conform to the edges of your original brain? That is, did the skull stripping remove everything outside of the brain while preserving the entire brain itself? Make sure to move your cursor around and inspect lots of different regions and boundaries of the brain.

• If you observe that some of the brain was cut away along with the skull, then your skull stripping was too aggressive. If you can still see parts of the skull or other non-brain tissue (e.g., meninges) then your skull stripping was too conservative. We can adjust this by manipulating the Fractional intensity threshold option on the BET GUI.
  • Larger numbers cut away more brain.
  • Smaller numbers cut away less brain.

• In the examples below I would not be satisfied with the results. (Remember, you can click on the images to see larger versions)
  • The image on the left (red overlay) was too aggressive and removed parts of the brain, specifically a large chunk of the frontal lobe.
    • I would therefore determine that the fractional intensity was too high, and run it again with a lower value.
  • The image on the left (blue overlay) was too conservative and left in areas outside of the brain.
    • I would therefore determine that the fractional intensity was too low, and run it again with a higher value.

2. Rerun BET with a different Fractional intensity threshold and then inspect the new results. Iterate this process until you are happy with your results. Tip: First try a large change in the Fractional intensity threshold, examine your result, and then split your original change in half, examine the new result, rinse and repeat.

Segmentation is the process of segmenting different anatomical features (i.e., separating them), including grey matter, white matter, and cerebral spinal fluid. FAST is a program that attempts to do this automatically.

1. Go back to the FSL GUI and select FAST.

2. This will bring up another GUI in which you provide the name of your skull stripped brain as the Input Image (using the File Selection box as before).

3. You need to provide an output base name. The default is the name of your skull stripped brain - which is usually appropriate.

4.You should also choose the Estimated Bias Field option by clicking on its box.

5. Start your segmentation by clicking Go

Once FAST is complete, you will have several additional files in your output folder. These include individual partial volume estimates for each tissue type where the voxel intensities represent (on a scale of 0 to 1) the estimated amount of each tissue type in each voxel.

FAST also creates 'hard' segmentations (the file with _seg in the file name) where each voxel is labeled by predominate tissue type.

In the example below I have loaded all three partial segmentation files as well as the hard segmentation file. Note the voxel values in the Location window.

• The seg_pve_0 file = 0.0 indicating that FAST estimated the voxel indicated by the crosshairs does not include any CSF.
• The seg_pve_1 file = 0.48 indicating that FAST estimated the voxel indicated by the crosshairs includes 48% grey matter.
• The seg_pve_2 file = 0.52 indicating that FAST estimated the voxel indicated by the crosshairs includes 52% white matter.
• The _seg = 2 indicating that the predominant tissue type in that voxel is greay matter.

Experiment with overlaying these results upon your original and skull stripped brain in FSLeyes.

We now have a pretty skull-stripped brain and two viewers to investigate them; FSLeyes for 2D “slices” and MRIcroGL for 3D renderings. Your final task is to use these tools to find several anatomical structures.