Using mrGray: Segmentation and Visualization

Heidi Baseler
created:  1/26/1999

Brian Wandell
updated:  8/22/2001

0.  Introduction.  This program, mrGray, can be used in a stand-alone mode to visualize data, segment gray matter, and produce flat maps of visual cortex.  Because it can be used in isolation, we distribute it and mrFlatMesh (along with source).  This page describes the basic uses of mrGray.  The documents in the unfolding software describe how to use the output from mrGray to produce flat maps.

1. Setup:  It is important to establish a directory structure where you will keep the anatomy data and the results of the mrGray analysis.  We use an directory structure that looks like this:

subjectName/anatomyFile.dat
subjectName/left/classprojects/date
subjectName/right/classprojects/date

We keep the project file, white matter and gray matter classification files inside this dated directory.  Then we keep the results of unfolding as sub-directories within this tree.

2. Run mrGray on a PC running Windows NT or Windows 95.  Under File menu in mrGray choose New project.   Assign it a name, and it will be saved as a file called: [projectname].mrp

3. Select anatomy data.  We usually call volume anatomy files vAnatomy.dat and identify them by the directory name (see above).  Once you have defined your project and selected the anatomy file, the data will be loaded and you will see an anatomical image.

4. Setting the visuals:


a. Under View menu, adjust Zoom - I usually use 4 or 5 when working with the occipital lobe.
 b. Under View menu, Set Greyscale.  This changes the visibility of the brain to the user, but it does not affect the automatic segmentation process.  However, you will find it extremely useful to have good visibility when judging the -segmentation, and when editing.
I usually find the contrast on a CRT works well with the following settings (or thereabouts): 
Brightness = 0.35, Contrast = 0.75, Gamma = 1.8
However, it will vary with your monitor type and anatomy data

5.   Setting colors for classification types
 (NOTE:  You will only have to do this the first time you run mrGray).

Under the File menu, choose Preferences.  Edit color for the white matter the default color lavender is not terribly visible. I suggest a dark purple instead.

6.      Set the volume of interest (VOI):

a.  Go to the Tools window and toggle on the VOI crossbars.
b.  Use the right mouse button to drag the dashed lines around the volume you would like to segment.  There are two VOI boundaries in each of the 3 dimensions; make sure you check all 3.

c. TIPS:
   i. The smaller the area, the faster the segmentation and editing process.  However, leave plenty of room for gray matter to grow.
 ii.  Flip between orientations (in the Tools window, or use Ctrl-O) to define  the VOI in 3d.
 iii.  To define one hemisphere, place the mid-sagittal VOI boundary as close to the midline as possible, but make sure not to cut off any of the hemisphere in the VOI.
iv.  Flip through all the slices in the VOI in all three orientations to make sure that all the gray matter will be included.

Here is an example of a VOI chosen to contain the left occipital lobe:

Sagittal view (red)

Axial view (green)

Coronal view (blue):

7.      Save project file:

 Under the File menu:  Save project.  This will update the project file and also save the classification file as  [projectname].class in the same directory as the project file, [projectname].mrp.  This file contains the classification of the data into white, CSF and unknown.  The gray matter has not yet been determined.  They will be determined and saved later.

8.      Automatic classification:

a.  Place position crossbars (solid red, green and blue lines) within the VOI using the right mouse button.
b.  Under Classification menu, Classify White & CSF.
c.  In MR Classify window:  Toggle on (click on) Preview and Auto Update
d.  Use the 6 slider bars to segment white matter and CSF to your liking.


TIPS:

i.  CSF Mean < Gray Mean < White Mean (by definition)
ii. Leave a generous amount of gray matter -- at least 3-4 layers  (pixels) of thickness between the white matter and CSF boundaries.
iii. To help fill in white matter:  Set the Noise Stdev higher than the default.  Set the Confidence up high, around 0.9.  Reduce the Smoothness to 3.  You can adjust these settings to your liking.

 

e.      Click on Apply and Exit
 f.  Flip through slices, reposition crossbars, and repeat above process.  Note that it is difficult to find one setting that will be suitable for all parts of the brain.  Therefore, you may have to settle for something part-way suitable, and then edit.

9.   Select connected white matter:

 a.  Back in the normal window, use Ctrl-left mouse click to select the white matter that is connected within the VOI.  Connected white matter turns red, while the unconnected white matter remains the original color.

 b.  Under Classification menu, Delete Unselected White, then Deselect White.

+

10.  Check gray matter:

 Under Classification menu:

 a.   Select Gray layers and make sure that it is set to 4.
 b.  Grow Gray Matter, and flip through slices in different orientations to check for areas where gray matter was missed.

  

        c.  Delete All Gray.

11.   Save project:  Under File menu, Save project.

12. Editing CSF:   (NOTE:  SAVE your project frequently during the editing process!)


 a.  Start in axial (green) orientation, and go to the top (first slice) of the VOI.
 b.  To prevent gray matter from intruding from the opposite hemisphere (the one you DON’T want to segment):  Go to Tools window, and select Edit mode: CSF and Edit size: 4.  Use the left mouse button to add a "wall" of CSF and block off the gray matter in the opposite hemisphere. 
If your gray scale is set well, you should be able to guess which gray matter belongs to which hemisphere. Another method, which is faster but harder to visualize, is to wall off the other hemisphere in just a few sagittal views: notice the traveling band of dark gray as you march through the sagittal views near the midline, and paint the dark band with CSF. Check how you did by looking through the axial views.
 c.  To prevent gray matter from growing inside ventricles (where it doesn’t exist):  Start at the first slice, go through each slice and look for/identify the lateral ventricle.  Fill it with CSF using the same Edit size and mode as in b.
d.  In some cases, you may need to build a wall of CSF between the cerebrum and the cerebellum -- if you delete all selected white matter (step 9, above) and find that you still get white matter in the cerebellum.

(NOTE:  The posterior horn of the lateral ventricle extends back into the occipital lobes and comes quite close to the fundus of the calcarine sulcus.  Therefore, it can accidentally connect to and become part of the calcarine, and cause problems in the unfolded representation.  You can see the posterior horn just to the left of the calcarine sulcus in the two images below on the right.)

Before steps b, c and d:

After steps b, c and d

 

13.   Editing White Matter:  (NOTE:  SAVE your project frequently during the editing process!)

 a.  Start in axial (green) orientation, and go to the top (first slice) of the VOI.
 b.  Go to Tools window, and select Edit mode: White and Edit size:  1.
c.  Use left mouse button to add white matter where needed.  Use right  mouse button to remove white matter where it extends too far into the gray  matter.  Remember to try and allow 3-4 layers of gray matter between white
and CSF.  This may require removing some CSF between sulci.  Switch Edit  mode to CSF and use right button to remove CSF.
 d.  Grow gray matter frequently to check work.  Delete all gray before  continuing the editing process; otherwise you can't add white matter or CSF  to an area that’s already classified as gray.
 e.  Repeat steps a-d in the other two orientations (sagittal and coronal).

Here's the same example as that used in steps 10 and 12 (left) above, after editing the white matter and CSF.  Notice that the gray matter is more complete:


 

14. Check for handles:  (NOTE:  If you’ve done a good job in editing, there should be very few handles (~10 or less)).

 a.  Under Classification window, choose Topology and Analysis.
 b.  In Topology window, click on Update.
 c.  Fill cavities and Delete satellite volumes, if there are any.
 d.  Click on Update Method 1 to find handles.
 e.  Double click on the coordinates for each handle, and the position bars in  the normal window will automatically go to the position of that handle.  However, you must click OK to get out of the topology window and go back into the normal window.

 

Here is an example of finding a handle and removing it by adding white matter:

 

f.  Use Alt-N and Alt-P to flip from one handle to the next or previous.
 g.  Flip back and forth through slices and orientations (Ctrl-O) to look for obvious gaps in the white matter (handles).  Ask experienced users for feedback.
 h.  Periodically run the Topology Analysis and Update to check how many handles you’ve removed.
 i.  When you are down to a few, stubborn handles that you can't figure out, click on the Update Method 2 button in the Topology window.  This method takes longer, but may give a better estimate of the position of the handle.

15.   Select connected white matter and save:

 When all handles are removed, and all editing is done, in the normal window, use Ctrl-left mouse click to select the white matter that is connected within the VOI.  Then, under Classification menu, Delete Unselected White, then Deselect White (just like in step 9b).  SAVE project!

16.   Select connected gray matter and save:

 Under Classification window, Grow gray matter one last time.  Use Ctrl-  left mouse click to select the connected GRAY matter this time.  The selected, connected gray matter will now turn blue.  Under the File menu, choose Save selected gray.  Name the gray matter file [projectname].gray, in the  same directory as the .class and .mrp files.

 

 

 

17.   You’re done!  Now that you have your gray matter file saved, you can render it in 3d.  Under the 3d menu, choose Build Visualization. 

 

The visualization mode has many options for lighting, orientation, and smoothing.  We believe it is important to have a good 3D rendering card (1999 price ~150-250 dollars) to make this part of the program fun.

 

18.  Final Comment.  We use mrGray in conjunction with a suite of tools for alignment, data analysis, and other computational tools.  To make full use of the program, it is good to know how to incorporate ROIs and data values on overlays.  These files are very simple to create (and thus inefficient).  You can also go on to unfold the gray matter surface using mrFlatMesh (in Matlab on the Unix system).

To learn more about the algorithms used in mrGray, and the companion program, mrFlatMesh, please review the following references:  And we wouldn’t mind an occasional citation of the work.  Thanks.

1. Teo, P.C., Sapiro, G. and Wandell, B.A. (1997). Creating connected representations of cortical gray matter for functional MRI visualization. IEEE Transactions on Medical Imaging 16: 852-863.

2. Visualization and Measurement of the Cortical Surface. 
B. Wandell, S. Chial and B. Backus (2000). Journal of Cognitive Neuroscience, vol. 12, no. 5. pp. 739-52