This is the protocol currently in use by EDEVEL group. (analyses: YALE, YALE2, leonard)

• Unix Commands
• Display program manual

Image Registration

Segmentation

• white matter
• gray matter
• csf
• background

Brain Masking

Useful terms

• native: native space of the scan - no scaling or orientation correction has been applied
• reslice: MNI space
• 305: Space that is the average of 305 individuals
• mr: MRI gray image
• seg: tissue segmented image - each tissue type (white matter, gray matter, CSF, and background) is composed of a different color
• rf: nu-corrected image
• brain: the skull stripped image
• tag: points picked for alignment or tissue classification
• mnc: file format - in order to open any file with "mnc" suffix, type Display and the name of the mnc file
• img: analyze format -used for BSE
• xfm: transformation file; contains parameters used to go from one space to another
• air: transform file to go from one space to 305 space

Image Registration

We are registering our brain with the average 305 brains in MNI space.  This registration is 3 translation x 3 rotation (no scaling).

1. Register the raw mr using the average brain and tags.

register /icbm/MNI_DATA/average_brain/norm_avg_305_mri_1mm_unfilt.mnc -tag /data/devel/Elizabeth/validity/norm_avg_305_mri_1mm_unfilt.tag case#_raw.mnc

2. Set the average brain to gray, and adjust the intensity of the raw mr to a comfortable intensity. (The image can be enlarged by holding shift + the middle mouse button, and it can be moved by holding shift + the left mouse button).

3. Pick the 10 tag points on the raw scan corresponding to the ten on the average brain. Be sure to always pick the points with the "record tags" icon located on the left of the screen. Never use the right mouse button, otherwise the tag points will not stay in order, and order is critical. This is especially important if you decide to re-pick a point. For example, if you decide to re-pick the first point, go back to the first point by moving the highlighted number up to 1 using the up arrow. Then when you re-pick the point, click the record tag icon. This will move your tag point while maintaining the order. NOTE: The very center pixel of the cross hair is the actual location of the tag point.

4. The ten tag points:

 A. Left and Right Cerebellum (1 & 2)

Scroll through the sagittal view. Keep an eye on the cerebellum, and watch for when it becomes so small that you can no longer see the horizontal striations. Then in the sagittal view, click on the very center of the cerebellum, seen as an oval shape, with the left mouse button (Fig 1, Fig 2).  In the coronal view you should be able to see a rough triangle formed by the transverse sinus (Fig 1, Fig 3).  Look at the coronal view, and select the tag by choosing the apex of the triangle formed by the transverse sinus, cerebrum, and cerebellum (Fig 2, Fig 4). Double check through the axial view that the tag point chosen is in approximately the same area as the average brain.

B. Corpus Callosum  (3 & 4)

Most Anterior Point:

In the sagittal view, scroll to the mid sagittal slice until you're at the most anterior point of the corpus callosum (Fig 5). Click on the most anterior point at which the corpus callosum curves. Be sure to look in all views of the raw scan and compare them to see how similar the image is with the averaged brains. Make sure you are in the very middle by checking in the coronal view that you are between the two hemispheres (Fig 5).

Most Posterior Point:

Again, go to the sagittal view. You must scroll through again because sometimes the head can be tilted to the left or right (which is visible in the axial view), and find the most posterior point of the corpus callosum (Fig 6). Click on the corpus callosum right before the point at which it starts to curve in, but after the round peak (Fig 6). Check in all views and compare your image with the sample images to double check. NOTE: be cautious of what you think the most posterior point is because sometimes the head can be tilted forward.

C. Left and Right Eye Sockets (5 & 6)

Scroll through the axial view until you can see the cornea quite distinctly. Pick the center of the eye socket (Fig 7, Fig 8). Scroll through the coronal view to make sure you are at the point where the diameter of the eye socket bones are the largest. Also make sure you are in the middle of the socket. Be sure to check all views, and make sure the tag point is centered in all the views (Fig 7, Fig 8).

D. Fourth Ventricle  (7)

Scroll in the mid sagittal view until you are at the very corner of the fourth ventricle (Fig 9). Look at the axial view to be sure you are in the middle and at the most posterior point of the fourth ventricle. If not just scroll through the sagittal view, while looking at the axial view, until you are at the most anterior point. Then pick the tag point (Fig 9).

E. Left and Right Temporal Lobe  (8 & 9)

Scroll through the coronal view until you are at the most anterior point of the left temporal lobe. NOTE: be aware of bone and fats that are right below the temporal lobe. Do not confuse the fats and bone with the temporal lobe. Click on the very center of the most anterior point of the temporal lobe. Be sure to pick in the slice right before it is no longer visible (Fig 10, Fig 11). Look at all views, and scroll through to be sure you are at the most anterior point. Again be cautious of fatty tissue and deposits. Scroll through and enlarge the image to be sure that you have picked the most anterior point (Fig 10, Fig 11). NOTE: be cautious of what you think is the most anterior point. You must account for the possibility of the head being tilted forward or backwards.

F.Mammillary Bodies  (10)

Go to the mid sagittal slice by scrolling through the sagittal view. Use the left and right arrows to go through slice by slice. Tag the mammillary body at the point where the body looks very round and clear. Another reference is to be able to see the brainstem, especially the pons, very clearly (Fig 12). NOTE: clear, meaning the structures look well defined, and have a very clear and rigid edge or boundary.

5. After selecting the tag points, go into the third window, and bring the blend bar all the way to the left. Now scroll through the brain and see: if the temporal lobes come in together, the ventricles come in and leave together, and if the parietal lobes come in together. If there is an asymmetry, check the other lobes and ventricles, and decide if the asymmetry is due to bad tag points, or actual asymmetries in the brain.  Also check to see if the midsagittal line is straight up and down.

6. Enter the name of the tag file in the top white box, and then press "save tags."

case#_reslice.tag

7. Enter the name of the transform file in the second white box, and then press the blank yellow box (to save)

case#_reslice.xfm

8. The reorientation transform file is now used to create a reoriented file.

mincresample -short case#_raw.mnc case#_reslice_raw.mnc -like /icbm/MNI_DATA/templates/icbm_template_1.00mm.mnc -transformation case#_reslice.xfm

Segmentation

Summary of tissue segmentation:

1. 120 points need to be tagged in the reslice_mr scan. When selecting the tag points, all three views need to be examined to make sure the tag is fully volumed. Although all views are used to view the tag point, it is only selected in the coronal view.  Move the cross hair by clicking with the left mouse button.  Select the point by clicking the right mouse button in the coronal view.  Try to take samples from posterior to anterior, and equally in both hemispheres.

2. Open the resliced gray-scale image in register.

register case#_reslice_mr.mnc

3.  The file opens in spectral color, change it to gray scale by selecting the "Gray" button.  Then adjust the contrast on the horizontal bar directly above that button to view it comfortably.  Fig 1

4.  Select the 120 tag points. Do not take any of the tag points in the cerebellum.

A. WHITE MATTER:

• Total of 40 samples to be taken.
• Label white matter samples "1"
• Do not take samples in the corpus callosum, in the anterior or posterior limbs of the internal capsule, or near the lateral ventricles.  Corpus Callosum: Fig 2, Fig 3 Ventricles: Fig 4a, Fig 4b
• Samples should tend to be taken more laterally in the hemispheres to avoid abnormal white matter that is frequently associated with the ventricles in the center of the brain.  Good points: Fig 5
• Samples should be evenly distributed across approximately 10 coronal sections, evenly spaced across the sagittal extent of the brain. Roughly 4 samples should be taken in each of the middle 7 or 8 sections.  Fig 6, Fig 7, Fig 8
• Take the required six samples first, then distribute the remaining 34 samples evenly across the brain beginning at the posterior extent and finishing at the anterior extent.

Required Samples:

• 1 left and 1 right hemisphere temporal lobe sample in the first section where the superior colliculi merge with the pulvinar nucleus of the thalamus.  Take the samples as lateral as possible to avoid abnormal white matter near ventricles commonly seen closer to midline in the temporal lobes. (Samples are not always taken in the same slice in the left and right hemispheres.)  Fig 9
• 1 left and 1 right hemisphere sample in the anterior 1/10th of the brain in the frontal poles.  Begin as anterior as possible and scroll posterior slice by slice until an adequate white matter sample can be taken. (Samples are not always taken in the same slice in the left and right hemispheres.)  Fig 10
• 1 left and 1 right hemisphere sample in the posterior 1/10th of the brain in the occipital poles.  Begin as posterior as possible and scroll anterior slice by slice until and adequate white matter sample can be taken. (Samples are not always taken in the same slice in the left and right hemispheres.)  Fig 11, Fig 12

General Guidelines:

Many samples in the middle 8/10ths of the brain will be taken in regions of the centrum semiovale extending from the parietal to the frontal lobes (points lateral to the cingulate cortex) Fig 13.  Many other samples should be taken superior to the lateral fissure--and parietal operculum--across the extent of the parietal and into the frontal lobe Fig 14, Fig 15.  Others more anterior to the lateral ventricles can be taken in the center of the frontal white matter Fig 16 and more laterally in each hemisphere closer to the cortical ribbon Fig 17.  It is important to check all three views for every sample taken to ensure that the tag point is entirely volumed white matter.  Some inadequate samples: Fig 18, Fig 19

After the 40th sample, type the number 1 in the far right window for later editing purposes.

B. GRAY MATTER:

• Total of 40 samples to be taken.
• Label gray matter samples "2".
• Take the required 12 subcortical samples first, and then distribute the remaining 28 cortical samples evenly across the posterior to anterior extent.
• Cortical samples should be evenly distributed across approximately 10 coronal sections evenly spaced across the sagittal extent of the brain. Roughly 3 to 4 samples should be taken in each of the middle 7 or 8 sections.  Various examples: Fig 20, Fig 21, and Fig 22
• Start taking cortical samples in the posterior 1/10th of the brain, taking the first in the most posterior section where the cortical gray ribbon can be clearly distinguished from white matter. In the most posterior sections, only one sample, unless you can find two good ones, should usually be taken in each hemisphere because there is less gray matter in those regions.
• Be especially careful to examine the sagittal and axial views to ensure fully volumed samples (excluding samples with partially volumed white and partially volumed CSF).  Also avoid samples in the very center of a gyrus where there is inevitably a sulcus containing CSF.  Some inadequate samples: Fig 23, Fig 24

Required Samples:

• Try to take these samples from posterior to anterior.
• The required tag points are listed by tissue type, but the numbers in parentheses reflect the order in which they are picked as you move posterior to anterior.

Thalamus:

• 1 left and 1 right hemisphere sample in the pulvinar of the thalamus. Take samples in the same section as the first white matter samples; where the superior colliculi merge with the pulvinar.  If the right and left sides merge in different sections, take both pulvinar gray matter samples in the more posterior of the two sections. NOTE: keep these samples relatively medial where the most robust gray signal can be found in this structure Fig 25. (1 & 2)
• 1 left and 1 right hemisphere sample more anteriorally in the dorsomedial nucleus of the thalamus: in roughly the center of the thalamus as viewed in the midsagittal plane in the section where the pulvinar sample was taken.  The appropriate slice is found by moving the curser to the apex of the thalamus in the midsagittal section Fig 26 and selecting right and left hemisphere thalamus points in that coronal section.  NOTE: keep these samples relatively medial where the most robust gray signal can be found in this structure Fig 27. (3 & 4)

Caudate:

• 1 left and 1 right hemisphere sample in the caudate nucleus in the coronal section where the anterior commissure crosses the midline (decussation) Fig 28. (5 & 6)
• 1 left and 1 right hemisphere sample in the caudate nucleus in the most anterior coronal section where the putamen is still present Fig 29. (11 & 12)

Putamen:

• 1 left and 1 right hemisphere sample in the putamen in the same section where the posterior caudate samples were taken: at the decussation of the anterior commissure . Take the sample more inferior and lateral where the most robust gray matter signal can be visualized Fig 30. (7 & 8)
• 1 left and 1 right hemisphere sample in the putamen more anteriorally, when the subcallosal gyrus can first be distinguished (by a white matter tract between the gyrus and the nucleus acumbens) from the basal forebrain gray matter. Take the sample in the center of the oval shaped form of the putamen at this level Fig 31. (9 & 10)

Continue taking the cortical gray matter samples.  After the 40th sample (tag number 80) type the number 2 in the far right window for later editing purposes.

C. CEREBROSPINAL FLUID SAMPLES:

• Total of 20 samples to be taken.
• Label CSF samples "3".
• The CSF samples should be, to the extent possible, evenly distributed from back to front, top to bottom, and left to right. The samples taken in the lateral ventricles should be, for the most part, taken from posterior to anterior.
• Take the first ten samples in the lateral ventricles (5 in each hemisphere), and the last ten in other subcortical CSF locations.
• Make sure not to take samples in the choroid plexus within the lateral ventricles Fig 32.
• Generally make sure to take samples in places where low signal values (dark voxels) can be seen in the center of the tag circle in all 3 orthogonal views (this will help avoid choroid plexus which has a signal value closer to gray matter) Fig 33, Fig 34.
• All of the samples should be taken in the lateral ventricles and the subcortical spaces e.g., the 4th ventricle, cerebral aqueduct, etc.; Fig 35, Fig 36, and Fig 37.

After the 20th sample (tag number 100) type the number 3 in the far right window for later editing purposes.

D. BACKGROUND SAMPLES:

• Total of 20 samples to be taken.
• Label background samples "4".
• Zoom out on the image for visualization of the entire background space Fig 38.
• Choose samples in approximately 5 coronal sections equally distributed across the coronal extent of the brain. Take samples from top left, top right, bottom left, and bottom right Fig 39.
• The signal value for each sample cannot be above 4.  This value might differ depending on scan intensity.  [If this is the case (as in the Leonard data) open the reslice_mr image in Display.  On the menu bar, select "S" Slice View menu and then "G" Recompute Histogram.  On the histogram that appears beside the contrast bar (to the left of the images), drag the bottom marker in line with the highest signal value in the background intensities peak Fig 40.  This value will be the maximum signal value for the background tag points.]
• No more than 4 of the 20 samples should be taken in the 0 background.

After the 20th sample (tag number 120), type the number 4 in the far right window.

4. Save the tag file by typing the filename in the white box between the "load tags" and "save tags" buttons.  When the name is typed, press enter then select the "save tags" button.  After saving the tags, quit register.

case#_reslice_seg.tag

5. The tag file now needs to be edited to fill in the 1s, 2s, 3s,and 4s that were not filled in.  Open the tag file in nedit, highlight all of the numbers until you see a "1" on the far right side.  Every row above the row containing the "1" should be highlighted.  Go to the search menu and select replace.  In the search for box type "", and in the replace with box type "1".  Select replace in selection, and there should now be ones on the far right of the first 40 lines.  Repeat this procedure for 2, 3, and 4.  Save the tag file.

nedit case#_reslice_seg.tag

6. The segmentation tags can be used to create the segmented image.

classify -verbose -min -tag case#_reslice_seg.tag case#_reslice_mr.mnc case#_reslice_seg.mnc

7. Create 2mm resliced files for the reslice_mr file and for the reslice_seg file to be used in brain masking.

mincresample -short case#_reslice_seg.mnc case#_reslice_seg_2mm.mnc -like /data/icbm/MNI_DATA/templates/icbm_template_2.00mm.mnc

mincresample -short case#_reslice_mr.mnc case#_reslice_mr_2mm.mnc -like /data/icbm/MNI_DATA/templates/icbm_template_2.00mm.mnc

8. Gzip the mnc files.

gzip *.mnc

Brain Masking

1. The reslice_mr, 2mm reslice_seg, and 2mm reslice_mr files are displayed to create the brain mask.  The mask will be drawn on the 2mm reslice_seg image.

Display case#_reslice_mr.mnc case#_reslice_seg_2mm.mnc case#_resliced_mr_2mm.mnc

2. Once display opens, there should be an image of a brain that is orangeish.  Type the following commands to set the paint label and masking threshold:

• "F"-to open the Segmenting menu
• "Y"-to set the Threshold
• Type "0.9 2.1" for the min and max threshold so as to only mask white and gray matter in the segmented image
• "D"-to set the Paint Label
• Type "4" for the current paint label (this is the color of the mask/label) - (This option also allows you to pick different colors for different ROIs and create different masks for each ROI by distinguishing between the paint label colors.)

3. The 2mm reslice_mr image will be altered to view it best.  Type:

• "Space bar"-to go to the main menu
• "D"-to open the Color Coding menu
• "D"-to change the image to Gray Scale
• Drag the bottom line of the color scale all the way down to zero, then adjust the top line so the image can be clearly viewed Fig 41.
• "C"-to change the Label Opacity
• Type "0.2" for the new label opacity (This opacity should be low so that the grey image can be seen underneath. Sometimes it is hard to distinguish between brain and non-brain, so the opacity can be lowered to better view the mr by making the label more transparent).

4. The 2mm reslice_seg image will be similarly altered.  Type:

• "Space bar"-to go to the main menu
• "S"-to open the Slice View menu
• "R"-to switch images (Repeat until the segmented image is in view)
• "Space bar"-to go to the main menu
• "D"-to open the Color Coding menu
• "S"-to change the image to Spectral
• Drag the bottom line of the color scale all the way down to the bottom.  Then adjust the top line (near 3.2) so the white matter is black, the gray matter is green, the fluid is red, and the background is white Fig 42.

5. The reslice_mr will now be adjusted.  Type:

• "Space bar"-to go to the main menu
• "S"-to open the Slice View menu
• "R"-to switch images (Repeat until the 1mm mr image is in view)
• "Space bar"-to go to the main menu
• "D"-to open the Color Coding menu
• "D"-to change the image to Gray Scale
• Drag the bottom line of the color scale all the way down to zero, then adjust the top line so the image can be clearly viewed Fig 43.

6. The brain mask can now be created.  Return to the segmented image view.  Type:

• "Space bar"-to go to the main menu
• "S"-to open the Slice View menu
• "R"-to switch images (Repeat until the segmented image is in view)
• "Space bar"-to go to the main menu
• "F"-to go the Segmenting menu

All of the work will be done on the segmented image in the coronal plane.  Pull the axis down so that the coronal view is at a comfortable size Fig 44. To adjust the image size, drag the mouse up and down on the image while holding shift + middle mouse button.  (Be sure caps lock is off.)  Holding shift + left mouse button will move the image in the window.  To navigate between sections/slices, use the "+" and "-" keys.

Use the right mouse button to paint the label.  Hold down shift + right mouse button to erase a label.  The brush size should be only one pixel for masking.  To change it, select "F" Segmenting menu, then "F" XY Radius. Type "1" for the new xy radius.

The whole idea of the brain mask is to fill in all that is brain, and exclude all that is not.  Often times that which is brain is touching that which is not.  To fill in only the brain, paint (using the right mouse button) the portion of the brain touching non-brain.  As soon as these "leaks" are filled in, move the curser on to the brain and press "E" Label Fill to fill in the brain.  The regions already painted and the CSF border will serve as a limit to how far the filling will go.  The threshold is set to include only gray and white matter, so the label fill will stop at the CSF border.  NOTE: It is only necessary to patch the leaks where the brain is touching non brain that segments as white or gray matter.  Fig 45: empty slice, Fig 46: holes filled in, Fig 47: slice masked

It is also difficult to distinguish brain from non-brain using only the segmented image.  This is the reason the mr images are displayed.  From the Slice View "S" menu, type "R" to toggle between images.  It is often easier to distinguish the brain/non-brain border in the mr, and then paint the correct brain pixels in the segmented image.

NOTE: Be careful when you paint the brain regions touching non-brain. If you miss even one pixel, the entire slide gets painted when you Label Fill.  Press 7 to Undo, and then fill in the missed pixels.

The masking process begins in the most anterior section in which brain tissue can be distinguished and continues for each slice of the segmented image.  Different regions of the brain have different areas that need to be avoided.  In the most anterior portion of the brain, avoid the dura mater between the two hemispheres Fig 48.  Then watch out for the meninges and superior sagittal sinus in between the two hemispheres Fig 49.  These will continue throughout the entire brain.  Everything in between the temporal lobes, from the pons forward, is non-brain Fig 50.  Sometimes the border between the temporal lobes and non-brain is very hard to distinguish.  If you aren't sure what's what, just leave it in, and ask later.  One of the more difficult things to distinguish in the segmented view is the optic nerve crossing.  The optic nerves emerge into the gap between the temporal lobes and the insula and then pass along side the temporal lobes Fig 54, Fig 55.  Use the mr in particular, to make sure these aren't included in the brain mask.  The next non-brain tissue to avoid is the transverse and sigmoid sinus separating the cerebrum from the cerebellum Fig 51.  In the most posterior regions, watch out for the dura mater in between the two hemispheres Fig 52.

When the brain stem appears, only label the part that is as inferior as the most inferior portion, bottom, of the cerebellum.  To do this, adjust the screen so the axial view can be seen.  Scroll through the slices until the most inferior portion of the cerebellum is visible.  Label the cerebellum and brain stem in this axial slice, creating an inferior border.  Return to the coronal view to finish masking the brain Fig 53.

When the brain is labeled, save the brain mask. To do this, type:

• "Space bar"-to go to the main menu
• "T"-to open the File menu
• "1"-to turn Crop Save Labels: OFF
• "W"-to Save Labels (brain mask) as a mnc

case#_reslice_mask_2mm.mnc

Please see Creating Cortical Objects protocol to use the output of the above described processes to obtain a surface extraction.

Back to Protocols.

If you have any questions, please e-mail:

esowell@loni.ucla.edu

Updated by Amy and Suzanne (5/02)