DTI/Callosal Segmentation/Cinch Tech


Jump to: navigation, search

[edit] Using CINCH for segmentation

CINCH is Dave Akers' tool for fiber analysis, it uses a 3d platform in which you can rapidly scroll through anatomy slices while viewing fiber tracts. CINCH works quite well for segmenting callosums, however, we do not use it in our callosal data analysis. This tutorial then, will mainly be useful for those who are using it as a means of parsing fiber groups.

First, read this quick article about the basics of CINCH:[1]

  • Note: Fibers are typically generated by custom scripts written in the lab (e.g. dtiTrackAllHemisphereFibers.m). The file output for these scripts are typically .m files, however, a new function (dtiWriteFibersPdb) allows fiber groups to be saved in .pdb format, which CINCH recognizes and opens readily.

Now that you are comfortable loading subject pathways, and pushing and pulling fibers, we can begin our parcellation.

Callosum segmented with CINCH.
An entire group, all fibers selected. Line on occipital fibers represents the line for selecting Occipital fibers.
The same group (from medial view), unselected; this is the group you will use to parse out all your fiber groups.

[edit] General method

CINCH conveniently allocates space for 8 separate groups, very nice, since callosal segmentation uses eight groups at a time. In CINCH, we do one hemisphere at a time, as opposed to switching back and forth as you do in mrDiffusion. Your style may vary, but we have found the following methods to work best:

  1. Load the fibers for one hemisphere.
  2. You'll notice that it is all blue, meaning it has all been selected, and also that the assignments are locked. Begin by de-selecting all the fibers, ('ctrl-shift-a'). The complete fiber group will turn grey.
  3. Now you can assign each group one by one without disturbing the other groups--because the assignments are locked. Select the number for which you will use to assign the groups. Click on the "one" (or whatever number group you are working on) box and use your selection mode of choice to grab the fibers you wish to group together. These fibers will change to the color of that number's fiber group.
    1. Now you must check to see if your fibers are accurate and precise (do this for each group).
    2. Currently, you have a large chunk of gray fibers and one blue fiber group. Begin by turning off the gray fibers (click on the "0"). This will leave your new group. Scroll through all the slices, watching carefully all the relevant sulcal and gyral landmarks. Does your group cross any boundaries? If it does, it is easy to fix, just change your mode to "Subtract" and take the offending fiber groups out. They will disappear, joining the gray, unassigned group.
    3. Now check to make sure you have all the fibers that belong to that group. Turn on your group of interest and the gray group. Now turn off you group of interest. Examine the fibers in the gray group that border your group. Did you miss any? If you did, just turn your group back on and add the ones you missed.
    4. If you have any other groups that you have already segmented that border this one, it is a good idea to turn them all on and scroll around through the slices, ensuring that all the fibers project to the correct locations.
    5. This completes the cleaning process. CINCH is quite powerful, so don't neglect to use all its tools, some groups are best viewed and segmented in the surface tools, some with the "match" and some with the "touch." Many problems will be unlocked by simply viewing the data from many angles. For the record, CINCH has an undo function (ctrl-z).

[edit] Callosal methods

This section deals with the methods specific to using CINCH to segment corpus callosums. It should be read after you read the mrDiffusion segmentation protocol so that you have an idea of the anatomical markers used to divide the groups.

We will work from the back of the brain to the front, except for the lateral frontal fibers which are last. Each group's selection follows the same general method as specified previously; here we delve into specific techniques for each group. The sulcal and gyral landmarks are exactly the same as when using mrDiffusion, which are specified by the Mori et. al. paper.

DTI tractography based parcellation of white matter: Application to the mid-sagittal morphology of corpus callosum
Hao Huang, Jiangyang Zhang, Hangyi Jiang, Setsu Wakana, Lidia Poetscher,
Michael I. Miller, Peter CM. van Zijl, Argye Hillis, Robert Wytik, Susumu Mori (2004)
NeuroImage, v. 26, pp. 195-205.

  1. Now we will go through group by group.
Occipital fiber group shown with the remaining unselected (grey) fibers.
    1. Occipital
      Begin by scrolling through the sagittal slices near to the mid-sagittal plane while examining the parieto-occipital sulcus. Now examine the fibers near this sulcus. Usually, there will be a clear split where parietal fibers leave the occipital fibers. It is easy to use the touch tool to add fibers by passing the pointer vertically over the occipital fibers just posterior to where the occipital and parietal fibers split. Scroll through again to make sure your selection is projecting properly. Clean as needed.
    2. Parietal
      1. Posterior Parietal Use the same gyral and sulcal geography as specified in the mrDiffusion segmentation paper to decide which parietal fibers are posterior. Select these using the touch tool. Now, and for the rest of the segmentation, you will have to get used to turning on and off the neighboring fibers and the unselected group. Once you have selected your PostPar fibers, you will want to view them with and without the occipital and unselected fibers present. This will help you ensure that you are not missing fibers and that none of your fibers cross the specified boundaries.
      2. Superior Parietal Turn on all three slices and flip the brain so that you are looking at the superior side of the axial slice. Move that slice to the top of the brain to expose the central sulcus. Line up the coronal slice with this sulcus--this is your marker for deciding whether or not fibers are parietal or frontal. Select the remaining parietal fibers with the touch tool. Clean as needed.
    3. Frontal
      Here you will need to identify the superior frontal gyrus (SFG). It is the "decider" for which fibers are superior or anterior.
      1. Superior Frontal Use the touch tool to select all fibers projecting to the superior frontal fibers. CINCH makes it easy to make judgments about looping frontal fibers that would be difficult to classify in mrDiffusion. Using the touch tool, sweep horizontally from the central sulcus and up toward the anterior boundary of the SFG. You may have to make additional passes at a more inferior level to catch the shorter fibers.
      2. Anterior Frontal Here you will have to use the same looping gyrus above the eyes to decide which fibers are orbital and which are anterior. Scroll through all the slices to get an idea. Then use the touch tool to select all frontal fibers between the anterior SFG boundary and the superior boundary of the orbital lobes.
    4. Orbital
      Using Cinch, it is easy to round up the orbital fibers. Basically anything inferior to the anterior frontal fibers will be orbital. Use your favorite method of selection to parse them together. Don't forget the ones that curve back to the infero-posterior portion of the orbital lobes.
    5. Temporal
      Move the coronal slice to the midpoint of the splenium of the corpus callosum. Switch to the surface tool and draw a polygon around the temporal fibers coming down across the tapetum--much in the same way that a temporal ROI is drawn. Once you have added the group, switch out to the touch tool and view your group's relationship to the corpus callosum. Remove any fibers that project from the inferior most section of the splenium.
    6. Lateral Frontal
      Turn on all the fiber groups and rotate the brain and fibers such that you are viewing it form the bird's-eye perspective. Check for fibers that protrude out of the main bundle to the lateral surface of the frontal lobes. Use the touch tool to select them into this final group.

Now that you have segmented all the fiber groups, turn them all off, except for the 0 group. Use it as your "not" condition to make sure you did not miss any fibers that should belong to one of the eight groups. Clean as needed. Now pull the groups back into mrDiffusion and save them according to the specified nomenclature. You will also have to re-assign the colors because CINCH's colors do not always move well into mrDiffusion.

Personal tools