DTI/Callosal Segmentation

From VISTA LAB WIKI

Jump to: navigation, search

Contents

[edit] Collosal Callosal Parcellation

This section describes the procedures for performing callosal parcellations.

This shows the fiber groups in a segmented callosum.

[edit] Background Reading

Before you begin parcellating callosums, you should read this 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.

The Huang and collegues paper provides our basic paradigm for performing parcellations. We follow their protocol for selecting the planes for drawing ROIs, as well as thier instructions for orienting the ROIs relative to meaningful sulcii and gyrii. You will want to have a copy of the paper on hand when you are parcellating, especially when you are gaining proficiency. When you are confused, consult the Huang paper first, then this page.


We primarily use mrDiffusion to do our callosal segmentations. We are also, however, beginning to implement Dave Akers' CINCH tool for the same process. The write up for using CINCH to do segmentation is here [1]. This page will also give you some general strategies for using CINCH, even if you are not interested in callosums.

[edit] Now we can start parcellating

The general procedure is to work from the back of the brain to the front of the brain, drawing each ROI in one hemisphere, then drawing the corresponding ROI in the other hemisphere. For instance, begin by drawing the Right Occipital ROI, then the Left Occipital ROI, then draw the Right Posterior Parietal ROI, then the Left Posterior Parietal ROI. This convention saves time finding the proper plane for each ROI. Once the plane is selected, you can simply draw all your ROIs (for that slice), instead of say, doing all of one hemisphere then the other, in which you will have to spend more time navigating through T1 slices.

You will make 16 ROIs (eight per hemisphere). It is critical to retain proper nomenclature for the ROIs, without it, the analysis scripts will not be able to "see" your hard work. Name the ROIs following this format exactly: Mori_HemisphereLocationLobe eg: Mori_RAntFront, Mori_LSupPar. Note: The Occipital, Temporal and Orbital lobes do not have a specified location, they are just listed as Mori_HemisphereLobe, eg: Mori_ROcc.

We will now go through each of the eight ROIs and then talk about cleaning and entering your ROI ratings in the callosum spreadsheet.

  1. Occipital ROIs
    This is an example of an occipital ROI
    File:Sagittal forParOcc Calc boundary marked (WinCE).PNG
    Red X shows the specified intersection
    1. Occipital ROIs are drawn in the Y plane where the parieto-occiptal sulcus intersects the calcarine fissure (see image at right).
    2. Scroll into your target hemisphere by starting in the mid-sagittal plane and moving, slice by slice, until it is clear where the two sulcii converge--viewing this convergence in the X image, or sagittal window. Set your new coordinates at the point of the intersection. Now shift your focus to the coronal slice (Y image) window.
    3. Now is a good time to view the fibers in mrMesh to see the path of the occipital trajectories. Leave the coronal slice "on" when you open mrMesh and use it to see what landmarks you should choose for drawing your ROI. Do this by looking closely at the superior-most boundary for occipital pathways, line that point up laterally with surrounding gyral geography. Now switch back into the surface view and find that gyral landmark on your coronal slice. That becomes (for now) the superior boundary for your occipital ROI.
    4. Time to draw the ROI. On the mrDiffusion menu bar, go: ROIs -> New Polygon -> Y image. Now draw your ROI on the coronal slice. Start on the medial surface of the hemisphere and click your way over, following any gyral geography, to that superior point that you found earlier. Click there, then click over to the lateral edge (again, following geography) and click your way around clockwise (counter clockwise on the left hemisphere) until you are back where you started. Right click when you are done to see your work. Most occipital ROIs look kind of like a mal-formed jack-in-the-box head, being an elliptical shape of some sort with a point sticking up to catch the superior occipital projections.
    5. These instructions are the same for each ROI and will be referenced by section number 1.5 in future ROI instructions. Now look at your fibers. Load the proper hemisphere's callosal fibers, then go:
      Menu bar -> Fibers -> And with current ROI. This will show all the fibers that connect your ROI to the callosal fiber group.
    6. General instructions cont. Use mrDiffusion to shift through all the slices where your fibers are projecting. In the case of the occipital fiber groups, you will want to watch the parieto-occipital sulcus very carefully. Shift through the slices and watch the dots move around, if any dots appear outside of the p-o boundary, your ROI is catching extra fibers. Update mrMesh to see where these fibers pass through your ROI. Fill your brain's visuospatial sketchpad with this information and trim your ROI as such, until no fibers pass outside of the specified sulcal or gyral landmarks.
  2. Parietal ROIs
    Parietal ROI
    1. Posterior Parietal. The posterior parietal ROIs are drawn in the same plane as the occipital ROIs and basically cover the superior half of the brain not covered by the occipital ROIs. Drawing these ROIs, then, is as simple as covering the rest of the brain in that slice. A good step here is to load all the hemisphere's fibers, then look at them with mrMesh. Observe the manner by which the parietal fibers are divided between superior and posterior projections. This distinction is among the toughest that you will make in the parcellation process. A good (but not always perect) rule is that the posterior fibers are limited to two gyri anterior of the par-occ sulcus. You will often see superior parietal fibers that pass through the plane of the posterior ROI; this is annoying. It is remedied usually by taking note of where the fibers touch the ROI and removing that section. Once you have drawn your ROI, follow sections 1.5 and 1.6 to clean errant fibers before moving to the next ROI.
    2. This section will also be repeated for all ROIs You have now drawn two neighboring ROIs and need to take care of fibers that may be shared between the two ROIs.
      1. Begin by loading the occipital fiber group. Now load the posterior parietal ROI. Go to the menu bar and select Fibers -> And with current ROI. MrDiffusion will now connect any fibers that are in the occipital group and pass through the parietal ROI.
      2. Move to the mid-sagittal plane. You will still have 2d fibers selected, so here you will immediately see any fibers that may pass through the callosum. Assuming you have some shared fibers, you now have to decide to which ROI they belong. Use mrMesh and scroll through the slices in mrDiffusion to see where the fibers project given what you know about sulcal and gyral boundaries.
      3. Once you have made a decision about where the fibers actually belong, you need to trim the other ROI such that the fibers no longer contact it. Sometimes you will not be able to separate all the fibers and you will have to compromise. Our scripts will delete any fibers shared between two ROIs, so losing one or two fibers will not greatly affect the data, however, more than 5 or 10 is worth nitpicking a little bit. This is the most time consuming part of segmentation; this must be carried out after each ROI is drawn to make sure that no ROI shares a significant number of fibers with another ROI.
    3. Superior Parietal.
      Superior Parietal ROI
      Drawing the Superior Parietal fibers requires shifting the plane of focus to the axial slices (Z image). Reference the Mori article and utilize his procedure for finding the proper axial plane. We add that you use his "omega" technique to find the correct slice, but also that it is a good idea to begin by putting the coronal slice at the plane of the central sulcus, then as you scroll down, you have a steadfast marker for deciding on the anterior boundary of your superior parietal ROI. Draw the ROI by starting at the medial edge of the central sulcus and tracing the sulcus with your mouse outward until you hit the lateral edge of the brain. Then sweep back and cover the posterior region of the axial slice, all the while tracing the gyral geography. Sometimes it is helpful to use the Posterior ROI as a boundary on the posterior side, to prevent the sharing of fibers.
  3. Frontal ROIs
    1. Superior Frontal. Superior frontal ROIs are drawn in the same plane as the superior parietal ROIs. Keep your marker at the central sulcus, use it as the starting point for drawing your superior frontal ROIs. Now you simply trace the anterior border of the central sulcus and swing it anterior, covering the rest of the superior frontal lobes (tracing gyral geography, of course). Check that this ROI does not share fibers with the superior parietal ROI.
    2. Anterior Frontal. To draw the Anterior Frontal ROIs, you must do another change of plane. Use the Mori paper to help you choose the proper plane--it is 1/3 the distance between the genu of the corpus callosum and the frontal poles. It is a good idea to view all the hemisphere fibers to help you choose the best plane to use; use the mid-saggital plane to see the superior frontal gyrus. Fibers projecting to the SFG are property of the sup. frontal ROI, even though they may appear to head anterior. Once you have chosen your plane, draw your ROI such that it envelopes the superior half of the slice; its inferior boundary will be the superior boundary of the orbital ROI. It may be a good idea to draw the orbital ROI first, then use it as the base for your anterior ROI. A common issue with the anterior and superior frontal fibers is the problem of looping fibers in the frontal lobes. Most of the time it is easy to see where these fibers belong, use the SFG and trim ROIs as needed. This is another area that may prove time consuming.
    3. Lateral Frontal. Move your axial slice to just superior of the corpus callosum. Switch the anatomy background to "Vector RGB." Use this to identify the corona radiata--which looks like a blue ellipsoid in the center of the brain. Move your sagittal slice to the lateral edge of this feature. Switch back to T1 and draw your ROI on the sagittal slice, beginning at the central sulcus and circling along the top of the brain slice, (minding gyral geography) down to the frontal pole and back up at an angle such to avoid any temporal fibers, but catching any lateral frontal fibers. This is another case where it helps to view the whole fiber group before drawing so you can see what the lateral fibers look like. Many brains will not have any lateral frontal fibers.
  4. Orbital ROIs.
    Orbital ROI
    1. The Orbital ROI can be a bit tricky. Use the sagittal slices on mrDiffusion to move through the slices next to the mid-sagittal plane. Look for the large looping gyrus above the eyes. This is the property of the orbital ROI.
    2. It may be difficult to judge the line between the anterior and orbital groups in the area of the frontal poles. Again, use the looping gyrus, anything superior to that belongs to the anterior group. It may also be of help to use the coronal slices and move posteriorly through the brain, usually you can see the differentation between the frontal polar and orbital fibers through a sulcus that becomes visible about 5 to 10 slices into the brain. This is the superior-most edge of the the orbital group.
    3. Again, check to see that the orbital ROI does not share any fibers with the anterior ROI.
  5. Temporal ROIs
    Temporal ROI
    1. Temporal ROIs are drawn by selecting the plane as specified by Mori--move the coronal slice to the middle of the splenium.
    2. Take a look at the whole fiber group using mrMesh. Make a note of any temporal fiber groups and try and get a feel for how they might look when you isolate them with an ROI, this will also help you draw your ROI effectively.
    3. Draw your ROI such that it extends up the tapetum (the section of white matter bordering the ventricle) and then have it move out into the temporal lobe which juts out laterally at this point in the brain.
    4. Fibers grown from temporal ROIs are often scraggly. Clean yours by making sure that no fibers come into the ROI from the inferior section of the splenium. It is normal for it to be a little rougher than other fiber groups, even after cleaning.

Make sure to follow all these instructions for both hemispheres, creating 16 total ROIs.

Now we have a rough sketch of the ROIs, it is time to clean them.

[edit] Cleaning ROIs

This is a "nots" group
  1. The "Not" condition. Here is where you see how many fibers you missed in your segmentation, it will also give you the chance to make sure that none of your ROIs share any fibers with another ROI. As you can see in the image, it is possible to get most of the fibers, but there will always be some that cannot be placed in an ROI without sacrificing other larger fiber groups.
    Here's how to do it:
    1. Save all your ROIs, then go to the menu bar: ROIs->Merge. A little box will pop up asking you which ROIs to merge. Select all the ROIs on one hemisphere and click okay. Your T1s will now be covered in a bunch of red lines (ROIs). Turn off your ROI and return to the mid-sagittal plane.
    2. Now load that hemisphere's complete fiber group. Go: Fibers-> Not with Current ROI. This process shows you all the fibers that you missed on your first round of segmenting. Many of the left over fibers will be short, little fibers that are easy to grab by selecting the nearest and most proper ROI as the current ROI, then move to the slice where the short groups are, and add polygons to your ROI until you have gotten all the short fibers.
    3. You will have some difficulty, however, in getting those fibers which are not obvious in their trajectory. Even worse, some fibers may be clear in their trajectory, but when you add them to an ROI's group, they will suddenly become shared by another ROI. You may have to remove a somewhat large section of ROI and replace it with a satellite in another location in order to gather the correct fibers. This is less elegant than gathering all the fibers with a single-bodied ROI, but it will work in a pinch.
    4. After each addition from the "Not" group into any ROI, make sure to double check that ROI's "And" status with its neighboring ROIs. After you have gathered all the fibers you can from the "Not" group and made sure you haven't made any extra shared fibers, do the merge step again and see what is left. The fibers now should be mostly thalamic and unreachable connections.


Gfox 18:21, 15 August 2006 (PDT)

Personal tools