FreeSurfer Tutorial

Cortical Reconstruction Step-by-Step

This document will guide you through the cortical surface reconstruction process, using the sample data set named "bert".

You will be guided, step by step, in generating a complete set of 3-D volumes and 2-D surfaces (including the gray/white boundary and the pial surface) for "bert".

The sample data set can be downloaded from:

http://www.nmr.mgh.harvard.edu/freesurfer

This software is distributed by the Massachusetts General Hospital NMR Center and CorTechs © 1999, 2000, 2001.

You will need software for Talairach registration of the image data to perform some parts of the reconstruction. This software is not distributed with FreeSurfer but can be downloaded from the Montreal Neurological Institute website:

http://www.bic.mni.mcgill.ca/software

Creating a New Subject

Cortical reconstruction begins by creating the directory structure where you’ll place the MR data files from the scanner and the reconstruction data files created by FreeSurfer.

Once FreeSurfer has been installed, type ‘csurf’ at the prompt to call up the csurf interface.

Select 'New Subject' under the File menu of the csurf interface.

Specify the path and directory where csurf should put the reconstruction data files, including the subject’s name (see next image). Name the directory appropriately for the subject (e.g. "bert" for the sample data set distributed with this tutorial).

Important: The subject shown in the images of this tutorial is named "bert". Bert’s data set is the same data set as test_subject3, released with the FreeSurfer distribution.

Click on ‘OK’ in the Choose Directory window:

Your new subject directory, called bert, has now been created. Several subdirectories were automatically created under the main directory. Cortical reconstruction data files will be automatically written to the appropriate directories during the reconstruction processes.

Importing the Subject’s Raw Data for Reconstruction

At this stage you should have bert’s raw MRI data, which you downloaded from the FreeSurfer website (see p. 1 for the URL). Now you are ready to import this raw MRI data into the appropriate directories. First, all the individual structural scan sessions or "runs" to be used in the reconstruction must be put into their own separate subdirectories in mri/orig, under a directory containing all the runs. The orig subdirectory can be reached through the path: ‘bert/mri/orig’. All MR data files must be moved to your subject’s orig directory using the UNIX or LINUX command window (see example to follow).

For example, bert has several subdirectories labelled 001, 002, and 003 under the mri/orig directory. These subdirectories contain the actual MR data for three structural scan runs. Looking inside any of these three directories, you should see the individual slice files for that particular run, e.g. I.001, I.002, etc., where "001" corresponds to the first individual slice for that series or run, etc.

e.g.

Go to your subject’s mri/orig directory:

cd /bert/mri/orig

Make three subdirectories here—one for each structural scan run:

mkdir 001

mkdir 002b

mkdir 003

Move the data from each of the three structural scan runs into their appropriate directory.

Once the original scanner data have been moved into their own subdirectories in the orig directory, the scans should be converted, motion corrected and averaged.

Preprocessing: Image Conversion, Motion Correction and Averaging

MR images must next be preprocessed to be useable for reconstruction.

First, the images must first be converted from the native scanner format to the COR format that is used by FreeSurfer. Next, if multiple structural scans are acquired, they must be corrected for motion, using AFNI binaries. Finally, multiple scans are averaged together to generate one motion-corrected data set to be reconstructed. The AFNI utilities are the copyright and work of Robert Cox, PhD at the Medical College of Wisconsin. For more information see:

http://varda.biophysics.mcw.edu/~cox/

NOTE: Conversion/averaging can be done from the csurf interface, or via the Unix/Linux command line. Motion correction must be done from the command line. See the following section for more details.

Conversion and Averaging with the csurf Interface

In csurf’s Subject Tools menu, select ‘Setup Structural Scans’.

You now will get a window showing all the data you just copied into your subject’s mri/orig directories. The next image shows your subject bert has three mri/orig directories, corresponding to its three structural runs: 001, 002, and 003.

By clicking on the first run (001), the window is further expanded. You should now see information about the type of image files in this directory: I.001.

Clicking READ HEADER gives you further information about the parameters of the image files. Any inaccurate information can and should be corrected here:

If you want to omit an acquisition that was previously selected, just reselect the directory and then click on ‘Unset Functional Dir’.

Click on CONVERT/AVERAGE to convert the MRI volume into the 256 coronal slices of COR format and average the acquisitions.

Do the same for any run directories you want to use for your reconstruction. Here you proceed to directory 002:

Again, by clicking READ HEADER you can view the parameters and make changes directly in the applicable fields, if necessary.

Here are the parameters for structural run 002:

Do the same for the last directory—click on CONVERT/AVERAGE:

A dialog box will alert you when the process is complete:

Each of the selected acquisitions is directly averaged without any motion correction. If motion correction is needed or desired, refer to the following section on motion correcting from the command line.

Converting, Motion Correcting and Averaging in the Command Line

1. Data Conversion from the Unix/Linux Command Line

Conversion of native MR data to COR format is performed using a process called mri_convert.

Enter the mri/orig directory in the Unix/Linux command line. This is where the raw MR data files are located.

Type in the Unix/Linux command line:

mri_convert <name_of_structural_run>/<first_MR_acquisition>

Where <name_of_structural_run> refers to the directory of the structural run to be converted and <first_MR_acquisition> refers to the first slice of the structural run.

For example:

mri_convert 002/I.001

Repeat this step for all the structural runs that will be motion corrected and averaged for reconstruction.

II. Motion Correction from the Unix/Linux Command Line

Motion correction of MR data is performed by using a process called register.csh.

Motion Correcting with register.csh

Enter the mri/orig directory in the Unix/Linux command line. This is where the raw MR data files are located.

Type in the Unix/Linux command line:

register.csh dir1 dir2 #dir3

Where dir1 refers to the first set of COR files and dir2 is the second set, which is rotated to match dir1. # is the scaling factor that the registration program uses. A value of 1 means no rescaling. 2 scales the images down by a factor of two and is less memory intensive than a factor of 1. dir3 is where the finished, motion corrected product goes

Repeat this step for all the structural runs that need to be motion corrected.

III. Averaging from the Unix/Linux Command Line

Averaging of motion corrected images can be done with the csurf interface by using the ‘Setup Structural Scans’ option, or from the Unix/Linux prompt by using the mri_average command.

Type in the Unix/Linux command line:

mri_average XXX XXX XXX

e.g.

mri_average 001 002 003

Talairaching

The next step is obtaining the talairach averages for the data. As indicated in the introduction, talairaching software is not distributed with FreeSurfer but can be downloaded from the Montreal Neurological Institute’s web page at:

http://www.bic.mni.mcgill.ca/software.

Starting the Reconstruction

Now that the MR data have been preprocessed, you can start the reconstruction using csurf.

Type on the Unix/Linux command line:

csurf

This will open the csurf display.

In the File menu, select ‘Open Subject’.

(Note: In future, when your subjects directory expands and you have multiple subjects, possibly in multiple directories, be sure to first set the subjects directory to the directory where your subjects’ data are by selecting ‘Set Subjects Directory’ in the Preferences menu (see below) and entering the directory path in the pop-up window. This ensures your subject’s name will appear in the subject: field.)

In the Preferences menu, select ‘View Logs’. This will expand the csurf window and allow information to be displayed under the VOLUME and SURFACE bars:

You can see how ‘view logs’ displays additional information in the expanded window:

Click on the small button labelled 'orig' and then click the 'VOLUME' bar to the left of it. This will bring up a TkMedit window displaying the orig volume.

orig volume

The orig volume is your raw MR data after preprocessing.

Process Volume

The next stage in the reconstruction process is Process Volume. This is a fully automated, 3-step procedure that first intensity normalizes the data, next removes the skull from the volume and last segments the white matter from the gray matter.

In the 'Subject Tools' menu, select the ‘Process Volume’.

This will bring up the Process Volume window. Click 'ProcessVolume'. The button will turn purple and read ‘QuitNORMALIZE’. This indicates that the procedure has started. Click on the purple button at any time if you wish to cancel the process. The three steps of the Process Volume procedure are listed in the Process Volume window.

IMPORTANT: Each individual step of Process Volume and the resulting volumes will be shown in sequence here, but you may prefer to let Process Volume run to its completion before you examine the volumes that it generates.

The first step of Process Volume is Normalize Intensities. This step takes approximately 20 minutes per hemisphere processor time. The output of the Normalize Intensities step is the T1 volume (intensity normalized volume). The T1 volume can be viewed in Tkmedit as for the orig volume (click on the ‘T1’ radio button and then the Volume button) when the Process Volume procedure is complete. A dialog box will inform you the sequence is finished:

The T1 volume is now ready to view:

The next step, Strip Skull, finds the skull and removes any skull and non-brain (eg. neck) structures. This step takes approximately 1 minute/hemisphere processor time. The output of the Strip Skull step is the brain volume. The brain volume can be viewed in TkMedit as for the orig and T1 volumes (click on the ‘brain’ radio button and then the Volume button) when the Strip Skull procedure is complete.

The brain volume is ready to view:

The final step of the Process Volume procedure, Segment White Matter, segments the white matter leaving the wm volume that can be viewed using tkmedit. This last step takes approximately 20 minutes/hemisphere.

The wm volume is now ready to view:

Create Surface

The next step in the reconstruction process is to create a surface from the volume using the 'Create Surface' procedure. The Create Surface procedure is a 6-part, automated process.

Select the Create Surface option under the 'Subject Tools' menu and then select 'Both' when asked which hemisphere you want to process.

This will bring up the ‘Create Surface’ progress window.

Click the ‘Create Surface’ button at the bottom of the processing window:

Fill white matter is now in progress.

The first step in Create Surface, ‘Fill White Matter’, separates the left and right hemispheres and separates the brain from the brain stem (below the pons). Each hemisphere is separately filled with a contiguous set of voxels. This step takes 3-10 minutes and results in the filled volume, which can be viewed in TkMedit by selecting the ‘filled’ button, and then clicking the VOLUME bar to the left of it:

The filled volume is now ready to view:

The next step in the Create Surface procedure, ‘Tessellate White Matter’, tessellates the voxels representing the left and right hemispheres. This step takes approximately 1 minute/hemisphere and results in the 2-D orig surface for each hemisphere, which can be viewed in surfer. Select the SURFACE bar in csurf to bring up the surfer tools, which include a graphics window where you can view the cortical surface:

Inflated orig surface, lateral aspect:

Choose 'orig' under the ‘surface:’ menu field to view the folded original surface with the surfer display:

The ‘smooth RH (and LH) white matter’ step smooths the initial surfaces generated (these surfaces are found in the rh.orig and lh.orig files for the right and left hemispheres, respectively). This step takes approximately 2 minutes/hemisphere processor time. Its output, the smoothwm surface, can be viewed using surfer.

Finally, ‘inflate RH (and LH)’, the last step of the Create Surface procedure, inflates the smoothwm surfaces (rh.smoothwm and lh.smoothwm) and ouputs the inflated (bold) image. The Inflate Surface step takes approximately 8 minutes/hemisphere processor time.

By this stage in the reconstruction process, you should have generated the following volumes and surfaces for your subject:

3-D volumes: orig, T1, brain, wm, and filled.

2-D surfaces: orig, folded, and inflated.

Your subject is now ready for functional overlay.

If you wish to conduct inter-subject averaging or morphological studies with your data, then you must proceed with the next steps of the reconstruction process.

Otherwise, your reconstruction is complete!

Editing Manual Defects

The next step involves manual editing of the segmentation procedure. This process is necessary because non-cortical structures adjacent to the cortex (e.g. fornix, basal ganglia, and lateral ventricle) can result in a cortical surface with large topological defects. A topologically correct surface is essential for generating a geometrically accurate spherical representation of the cortex. Depending on the quality of the data set, this step takes approximately an hour, and may need to be repeated to get a surface free of large defects. Small defects will be automatically fixed in the subsequent step, ‘Fix Topology’.

The next three sections provide brief overviews of:

1. the TkMedit tools (‘Overview of the TkMedit Tools’),
2. the Surfer tools (‘Overview of the Surfer Tools’), and
3. the defect editing process (‘Overview of Defect Editing’).

They are followed by the ‘Editing Defects Tutorial’, which is a detailed guide to correcting defects on the sample data set.

You may wish to briefly familiarize yourself with the tools and review the defect editing process before proceeding to the Editing Defects Tutorial. You can call up the TkMedit and Surfer tools for your subject so they’re available while you skim through the overviews. To do so, select your subject in the ‘subject:’ field of csurf. Then go to the Subject Tools menu, and click on ‘Edit Segmentation’.

‘Edit Segmentation’ by default will bring up both your subject’s wm volume in the TkMedit graphic window, and the inflated surface in the Surfer graphic window.

Overview of Defect Editing

• Go to the File drop-down menu.

• Select Load Aux Volume; this generates a pop-up where you can load a second volume. Type "T1" in the field. Click OK button.
• When OK pops back up, you can view the second volume.

• To toggle the two volumes displayed, go to the Display menu. Select Aux Volume, or type <ALT> <C>. This will allow you to view and compare the segmented volume (wm) with the original MRI volume (T1) to determine how the defect needs to be fixed.

• Save the location of the point by clicking SAVE PT in the surfer window. This will allow you to view the selected point in the medit graphics window (see next step).

• Go to the point in the volume by clicking Goto Point on the TkMedit interface.

5) You’ll re-run Create Surface to generate a new surface affected by your changes.

In csurf, make sure your subject’s name is selected in the ‘subject:’ field. Then go to the Subject Tools menu, and select ‘Edit Segmentation’.

‘Edit Segmentation’ will by default bring up both your subject’s wm volume in the TkMedit Tools graphic window, and the inflated surface in the Surfer Tools graphic window. You will be manually editing the wm volume itself with the TkMedit tools, and using the surfer tools to refer to the topological quality of the inflated surface.

TkMedit reads a maximum of two volumes at a time. The images in the first buffer can be edited and saved, whereas images loaded into the second buffer are view-only and cannot be edited. The second buffer is useful for comparative analyses, (e.g. when the first buffer has segmented white matter (wm volume) and the second buffer has the original T1 images). It is recommended that you load in the T1 volume as your second volume (auxiliary to the wm volume) to refer to when you make your edits.

File, Edit, Display, and Tools.

Go to the ‘File’ menu on the TkMedit Tools interface:

Choose the first item, ‘Load Aux Volume’.

A small window will pop up. Enter ‘T1’ and click ‘OK’:

IMPORTANT: Be sure to type ‘T1’ with an upper case "T".

When the OK button pops back up, you are ready to view the auxiliary volume. You can toggle between the main and auxiliary volume by pressing ALT-C, or by alternately selecting the volume you want (‘Main Volume’ or ‘Aux Volume’) in the ‘Display’ menu:

Next, select ‘Edit Voxels’ in the ‘Tools’ menu:

Now you are ready to edit voxels and display your main (wm) and auxiliary (T1) volumes.

MIDDLE-CLICK and drag: to draw/fill voxels

RIGHT-CLICK and drag: to erase voxels

To undo your last edit, go to the ‘Edit’ menu and select ‘Undo Last Edit’:

This is what you should see in the Surfer window after selecting Edit Segmentation: the lateral aspect of the right hemisphere, with sulci and gyri indicated in red and green, respectively:

To switch the coloring off, thereby making visualization of defects easier, select the ‘none’ button to the right of ‘Background Color’ on the surfer interface, then click on REDRAW.

Now, to rotate the hemisphere so that the medial aspect is showing (where most of the defects are located), slide the ‘ROTATE (deg)’ bar to the farthest left (180º). Click REDRAW again. Your inflated surface should now look like this:

The objective is to remove all the topological defects (handles and holes) on your inflated surface, so that it can eventually be transformed into a sphere. It needs to look like this:

Editing Defects

This procedure follows the Create Surface step, once the surface has finished inflating. This step is only required if you wish to proceed with inter-subject averaging and morphological studies. It is also the only manual step of the whole reconstruction process and may require a few iterations.

To generate the final surfaces (gray/white boundary and gray/csf boundary), flatten cortical patches, or register the cortical surface, the cortical surface must be topologically equivalent to a sphere. In other words, there cannot be any holes or handles in the cortical surface that would prevent the hemisphere from looking like a smooth sphere if its surface were inflated. The initial estimation of the cortical surface usually has some topologic defects. There are two main causes of topologic defects:

1. segmentation errors (i.e. a white matter voxel is incorrectly classified as non-white or vice-versa)
2. non-cortical anatomic structures (e.g. fornix, basal ganglia, ventricles)

Typically, segmentation errors result in small (in area) topologic defects, while the non-cortical structures result in large topologic defects. The small topologic defects can be automatically fixed (see ‘Topology Fixing’). However, before the automated fixing can proceed, the larger topologic defects that arise from non-cortical structures must currently be manually corrected.

The most common anatomically derived defects that require manual editing are:

1. Fornix (results in handles in the hemisphere’s surface)
2. Lateral Ventricle (results in a hole in the hemisphere’s surface)
3. Basal Ganglia (results in a hole in the hemisphere’s surface)
4. Optic Nerve (which doesn't result in topological defects, but interferes with inflation)

A description of these defects and suggested manual editing procedures are presented in the next section.

Inflated Surface Showing Typical Topological Defects

(right hemisphere—medial aspect)

Guidelines for correcting each of these defects follow.

To view the corresponding defects in the wm volume:

1. In surfer, point your cursor to any defect and LEFT-CLICK. The selected point should now appear bright blue on the surface.
2. Now click on the ‘SAVE PT’ button on the surfer tools interface.
3. When the ‘SAVE PT’ button in surfer pops back up, click on the ‘Goto Point’ button in the TkMedit tools interface.
4. Your cursor in the TkMedit graphic window will now be on the point in the 3-D volume that corresponds exactly to the point you just selected on the 2-D surface.

1) LATERAL VENTRICLE

The lateral ventricle appears as an indentation on the inflated surface, and requires filling. Compare this surface to that shown at the end of the Lateral Ventricle section

Where To Start

Start filling in when the ventricles appear, and keep filling each ventricle until it is no longer enclosed. The coronal view and a brush of radius 1 or 2 are best for correcting lateral ventricle defects. The lateral ventricle needs to be filled in both hemispheres. Use the following images to guide you as you edit the lateral ventricle defects in the right hemisphere. The region requiring correction is labeled with a white circle.

Be sure to repeat the process in the left hemisphere.

LATERAL VENTRICLE—coronal slice 48

You can see the beginnings of the lateral ventricle in this slice, but no hole to fill yet:

LATERAL VENTRICLE—coronal slice 48

wm zoomed in T1 zoomed in

TIPS:

Press ALT-C to toggle between wm (main) volume, and T1 (auxiliary) volume, or go to the ‘Display’ menu in TkMedit and alternately click ‘Main Volume’ and ‘Auxiliary Volume’.

Use the ZOOM bar on the TkMedit Tools to zoom in and out.

LATERAL VENTRICLE—coronal slice 49

T1 volume

Scrolling forward to the next slice, you clearly see a hole that needs to be filled. If this hemisphere were to be inflated without filling the hole, you can imagine how it would resemble a doughnut, rather than the sphere required for subsequent processing. Back in the wm volume, point your cursor to the same area shown, then MIDDLE-CLICK & drag your mouse to fill in the hole completely as in the corrected slice below:

BEFORE CORRECTION AFTER CORRECTION

TIP: To undo an edit, go to TkMedit’s ’Edit’ menu and select ‘Undo Last Edit’.

LATERAL VENTRICLE—coronal slice 50

BEFORE CORRECTION AFTER CORRECTION

You clearly see another slice where the ventricle is completely surrounded by white matter, thereby resembling a hole in middle of this hemisphere’s surface. Continue to fill in the lateral ventricle as you scroll forward through the coronal slices.

LATERAL VENTRICLE—coronal slice 51

BEFORE CORRECTION AFTER CORRECTION

TIP: Use the ‘Brush Size’ option in TkMedit’s ‘Tools’ menu to increase your brush size as the area to be corrected increases in size.

LATERAL VENTRICLE—coronal slice 52

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the lateral ventricle in the wm volume, zooming in to help you see the defect. You may opt to fill in the lateral ventricle of the other hemisphere at the same time, when you become more proficient. The disadvantage of doing so is that TkMedit can only load the orig surface (overlaid in yellow) for one hemisphere at a time, making it sometimes hard to identify the outline of the orig surface (what has been segmented as wm) in the hemisphere that doesn’t have its orig surface overlaid.

TIP: To load the orig surface of the other hemisphere during edits, select ‘Load Surface’ from TkMedit’s File menu.

Next, type the appropriate hemisphere/surface (e.g. "lh.orig" for the orig surface of the left hemisphere) in the field of the pop-up window. All available surfaces can be displayed by selecting them in the TkMedit’s ‘Display’ menu.

LATERAL VENTRICLE—coronal slice 53

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 54

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 55

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

TIP: At any time, you can save the changes you’ve made to the wm volume by selecting ‘Save Volume’ in TkMedit’s ‘File’ menu, and clicking ‘OK’:

LATERAL VENTRICLE—coronal slice 56

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 57

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 58

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 59

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 60

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 61

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 62

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 63

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 64

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 65

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 66

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 67

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 68

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 69

BEFORE CORRECTION AFTER CORRECTION

TIP: Sometimes a subject’s lateral ventricles may appear enclosed in some slices (i.e. appear as holes), open in subsequent slices (i.e. no longer appear as holes), then enclosed again as you scroll forward. The rule of thumb when editing these is to keep filling until you reach the slice where they finally open up and are no longer enclosed.

LATERAL VENTRICLE—coronal slice 70

BEFORE CORRECTION AFTER CORRECTION

Here is an example of ventricles being opened up. In subsequent slices (see slice 78), they are enclosed again, so you must continue filling. If it were truly the last slice in which they were opened up (usually anteriorly—see slice 92 on p. 68), you could stop the edits on the lateral ventricle.

LATERAL VENTRICLE—coronal slice 71

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 72

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 73

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 74

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 75

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 76

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 77

BEFORE CORRECTION AFTER CORRECTION

 

 

LATERAL VENTRICLE—coronal slice 78

BEFORE CORRECTION AFTER CORRECTION

In slice 78, you see that the ventricle is enclosed again.

LATERAL VENTRICLE—coronal slice 79

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 80

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 81

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 82

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 83

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 84

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 85

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 86

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 87

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 88

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the ventricle in the wm volume.

LATERAL VENTRICLE—coronal slice 89

BEFORE CORRECTION AFTER CORRECTION

LATERAL VENTRICLE—coronal slice 90

BEFORE CORRECTION AFTER CORRECTION

These are the last few slices where bert’s ventricle is enclosed and needs filling.

LATERAL VENTRICLE—coronal slice 91

BEFORE CORRECTION AFTER CORRECTION

TIP: When editing the lateral ventricle in the future, you may find it easier to edit the first slice (e.g. bert’s 49) and last slice (e.g. bert’s 91) first, thereby "capping" the slices that need ventricle filling. Then it is just a matter of filling all the slices in between.

LATERAL VENTRICLE—coronal slice 92

wm volume T1 volume

Here you see that the ventricle is no longer enclosed, so you stop the filling.

Sometimes there is variation between a subject’s left and right hemispheres, so that in a particular slice one hemisphere’s ventricle will finally "open up", but the other hemisphere’s ventricle is still enclosed. In such cases, continue to fill the enclosed ventricle only, even though the other is open. This will address the topological defect.

After editing the ventricle, don’t forget to save the changes you’ve made to the wm volume, by going to ‘Save Volume’ in TkMedit’s ‘File’ menu, and clicking ‘OK’.

Inflated Surface Shown After Lateral Ventricle Defect Corrected:

This image shows how your edits to the lateral ventricle affect the inflated surface. It was generated by running ‘Create Surface’ after saving the edits made to the lateral ventricle in the wm volume.

It is used for illustrative purposes only—to save time, you may prefer to edit all the defects before running ‘Create Surface’.

See the next page for instructions on generating a new surface after your edits to the volume are saved.

1) In csurf’s ‘Subject Tools’ menu, select ‘Create Surface’, then select appropriate hemisphere(s) to process:

2) Select ‘Create Surface’ in the box, where you can then monitor its progress:

3) An ALERT box will inform you when the ‘Create Surface’ process is complete:

2) BASAL GANGLIA (Caudate, Putamen)

The basal ganglia defect, which needs filling, is often best seen on the surface’s lateral aspect. Compare this surface to that shown at the end of this defect’s editing section.

Where To Start

To start the edits, find the first slice where the outline of the basal ganglia appears as an indentation or "bay" on the white matter surface. A sagittal slice located between the basal ganglia and the insula is usually a good place to start filling. Continue filling until the basal ganglia no longer appear patchy. Corrections are best made in the sagittal view (although they can be made in the horizontal or coronal view as well) and the optimum brush size is 1 or 2. Use the following images as your guide to correcting the basal ganglia in the right hemisphere.

When finished, be sure to repeat the process in the left hemisphere.

BASAL GANGLIA—sagittal slice 116

This is the first slice where you see the beginnings of the basal ganglia defect, but because the strand beneath the cursor (red crosshair) isn't joined to the anterior portion of the cortex yet, it wouldn't cause a topological defect and doesn’t yet need filling.

Compare with slice 115.

BASAL GANGLIA—sagittal slice 116

wm zoomed in T1 zoomed in

BASAL GANGLIA—sagittal slice 115

BEFORE CORRECTION AFTER CORRECTION

Moving laterally to slice 115, you see how the strand connects to the anterior aspect of the cortex, resulting in a problematic indentation. This is where to start the filling.

BASAL GANGLIA—sagittal slice 114

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 113

BEFORE CORRECTION AFTER CORRECTION

In slice 113 you see that the strand above the cursor no longer connects to the anterior aspect of the cortex, however you should continue to fill in the area because in subsequent slices the strand reappears. As with the ventricle defect, once any edits start they should continue until the defect no longer appears in any subsequent slices.

BASAL GANGLIA—sagittal slice 112

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 111

BEFORE CORRECTION AFTER CORRECTION

In these sagittal slices you can clearly notice the filling in the lateral ventricle, resulting from the previous edits made in the coronal view.

Continue to fill in the basal ganglia.

BASAL GANGLIA—sagittal slice 109

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 108

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the basal ganglia.

BASAL GANGLIA—sagittal slice 107

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 106

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the basal ganglia. You may notice that the small holes to the left of the basal ganglia (see arrow) have been left unfilled in some slices, and filled in others. Filling these holes is unnecessary, and filling them by accident will have no impact on the surface—they are not segmented as holes in any case, as indicated by the lack of yellow (orig) surface overlay outlining their perimeters.

BASAL GANGLIA—sagittal slice 105

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 104

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the basal ganglia.

BASAL GANGLIA—sagittal slice 103

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 102

BEFORE CORRECTION AFTER CORRECTION

Notice again that we didn’t fill in the small, patchy holes surrounding bert’s filled basal ganglia. This is because they are already contained within the orig surface (overlaid in yellow), and therefore won’t result in topological defects. This is a good example of why it is preferable to edit defects in the hemisphere that has its orig surface overlaid.

BASAL GANGLIA—sagittal slice 101

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 100

BEFORE CORRECTION AFTER CORRECTION

Continue to fill in the basal ganglia.

BASAL GANGLIA—sagittal slice 98

BEFORE CORRECTION AFTER CORRECTION

BASAL GANGLIA—sagittal slice 97

BEFORE CORRECTION AFTER CORRECTION

Above you see the last slice where the basal ganglia needed to be filled, since in the next slice the area opens up. Below you see the coronal view of the filled basal ganglia.

BASAL GANGLIA—coronal slice 125

.

Be sure to save the changes you’ve made to the volume by selecting ‘Save Volume’ in TkMedit’s ‘File’ menu, and clicking on ‘OK’.

Inflated Surface Shown After Basal Ganglia Defect Corrected:

This image shows how your edits to the basal ganglia affect the inflated surface. It was generated by running ‘Create Surface’ after saving the edits just made to the basal ganglia in the wm volume.

It is used for illustrative purposes only—to save time, you may prefer to edit all the defects before running ‘Create Surface’.

3) FORNIX

The fornix appears as raised connections or "handles" on the inflated cortical surface and needs to be erased. Compare this surface to that shown at the end of the Fornix editing section.

Where to Start

To start erasing the fornix, find the first lateral slice where the fornix is present and erase carefully. The erasing process is best done in the sagittal view using a brush size of 1. Continue to erase the fornix in all subsequent slices, including where the fornix connects the temporal strand to the noncortical structures. Use the following images as your guide to correcting the fornix in the right hemisphere.

When finished, be sure to repeat the process in the left hemisphere.

FORNIX—sagittal slice 105

wm volume T1 volume

This slice clearly shows the ventricle and basal ganglia fills already made. You also see the beginnings of the fornix. It appears here as a projection (see area to the lower left of the cursor) and in subsequent slices will appear as a distinct strand or handle. Start erasing the fornix as soon as its projection appears (edit of this slice follows).

FORNIX—sagittal slice 105

wm zoomed in T1 zoomed in

FORNIX—sagittal slice 105

BEFORE CORRECTION AFTER CORRECTION

Here is where to make the first edit of bert’s fornix.

FORNIX—sagittal slice 106

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 107

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 108

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 109

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 110

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 111

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 112

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 113

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 114

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 115

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 116

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 117

BEFORE CORRECTION AFTER CORRECTION

In these slices the fornix appears as a small chunk of wm, and still needs to be erased.

FORNIX—sagittal slice 118

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 119

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 120

BEFORE CORRECTION AFTER

FORNIX—sagittal slice 121

BEFORE CORRECTION AFTER CORRECTION

In these slices, the fornix once again forms a strand. Continue to erase it, being careful not to erase any adjacent cortical structures.

FORNIX—sagittal slice 122

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 123

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix.

FORNIX—sagittal slice 124

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 125

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix along the entire length of the column.

FORNIX—sagittal slice 126

BEFORE CORRECTION AFTER CORRECTION

FORNIX—sagittal slice 127

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the fornix for these last two slices. In the slice below you notice the absence of the yellow orig surface overlay, because you are at the mid-point of the brain (128 of 256 slices) and moving out of the right hemisphere. You would now have to load the left hemisphere’s orig surface to optimally edit that hemisphere.

FORNIX—sagittal slice 128

BEFORE CORRECTION AFTER CORRECTION

This is your last edit to the fornix of the right hemisphere. Save your changes to the volume by selecting ‘Save Volume’ in TkMedit’s ‘File’ menu, and clicking ‘OK’.

Inflated Surface Shown After Fornix Defect Corrected:

This image shows how your edits to the fornix affect the inflated surface. It was generated by running ‘Create Surface’ after saving the edits made to the fornix in the wm volume.

It is used for illustrative purposes only—you may prefer to edit all the defects before running ‘Create Surface’.

IMPORTANT: Notice how a small handle still appears on the surface (see arrow). This means a defect wasn’t edited, and needs to be addressed. Click on the defective area on the surface so that it is highlighted in bright blue. Then click on ‘SEND PT’ on the surfer tools. When the button pops back up, click on ‘Goto Point’ in the TkMedit tools—your cursor should now be pointed directly on the defect in the wm volume in TkMedit. Investigate carefully, and edit again. If editing doesn't work, the next step (automatic topology fixer) will likely address the defect.

4) OPTIC NERVE

The optic nerve appears as a sharp projection on the inflated surface, and although it doesn't result in a topological defect, it should be removed. The high curvature of this non-cortical structure unnecessarily complicates flattening and spherical morphing. Compare this surface to that shown at the end of the Optic Nerve defect editing section.

Where To Start

Find the first sagittal slice where the optic nerve is present, and erase it. Continue to do the same when scrolling through each subsequent slice until the optic nerve is no longer visible. A brush size of 1 works well for this defect. Comparing the T1 image is useful in slices where the optic nerve looks fragmented. Use the following images as your guide to correcting the optic nerve in the left hemisphere.

When finished, be sure to repeat the process in the right hemisphere.

OPTIC NERVE—sagittal slice 116

wm volume T1 volume

This is the first slice where you see the beginnings of the projection of the optic nerve.

See how it should be edited on the next page.

OPTIC NERVE—sagittal slice 116

wm zoomed in T1 zoomed in

OPTIC NERVE—sagittal slice 116

BEFORE CORRECTION AFTER CORRECTION

Slice 116 (above) requires minimal correction. Slice 117 (below) requires slightly more. Be careful not to erase the cortical structure when editing the optic nerve.

OPTIC NERVE—sagittal slice 117

BEFORE CORRECTION AFTER CORRECTION

OPTIC NERVE—sagittal slice 118

BEFORE CORRECTION AFTER CORRECTION

Here is a close-up of slice 118, AFTER CORRECTION:

OPTIC NERVE—sagittal slice 119

BEFORE CORRECTION AFTER CORRECTION

Scrolling medially, the optic nerve defect becomes more obvious. Continue to erase it.

OPTIC NERVE—sagittal slice 120

BEFORE CORRECTION AFTER CORRECTION

OPTIC NERVE—sagittal slice 121

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the optic nerve defect.

OPTIC NERVE—sagittal slice 122

BEFORE CORRECTION AFTER CORRECTION

OPTIC NERVE—sagittal slice 123

BEFORE CORRECTION AFTER CORRECTION

Continue to erase the optic nerve defect.

OPTIC NERVE—sagittal slice 124

BEFORE CORRECTION AFTER CORRECTION

OPTIC NERVE—sagittal slice 125

BEFORE CORRECTION AFTER CORRECTION

You can stop your edits in the slice below because the optic nerve has no orig surface overlaid in yellow, and is therefore no longer identified as wm in this hemisphere:

OPTIC NERVE—sagittal slice 126

Inflated Surface Shown After Optic Nerve Defect Corrected:

This image shows how your edits to the optic nerve affect the inflated surface (see arrow). It was generated by running ‘Create Surface’ after saving the edits made to the fornix in the wm volume.

IMPORTANT: Notice how a small handle still appears on the surface (see arrow). This means a defect remains, and needs to be addressed. Click on the defective area on the surface so that it is highlighted in bright blue. Then click on ‘SEND PT’ on the surfer tools. When the button pops back up, click on ‘Goto Point’ in the TkMedit tools—your cursor should now be pointed directly on the defect in the wm volume in TkMedit. Investigate carefully, and edit again. If you can see no connected voxels to account for the defect, then the automatic topology fixer should be able to address it. (see the Fixing Topology section for more information).

Once the surface finally looks to be free of defects, you can proceed to the next step, the fully automated ‘Fix Topology’ process.

Additional Topological Defect: The Temporal Lobe

After initial segmentation, many subjects have an additional topological defect on their inflated surface. In these cases, the temporal lobe defect results from the weak intensity obtained from the thin strands of temporal gyri, and it is addressed by thickening the strand(s).

This subject's inflated surface after initial segmentation is showing the temporal lobe defect, which usually presents as a handle on the medial aspect. The temporal lobe defect below has been highlighted with the cursor.

The corresponding point in the 3-D volume is shown on the next page, along with editing instructions.

Here is the corresponding point in the T1 volume, with the orig surface overlaid:

You can see how the orig surface of the gyrus' thinnest strand is fragmented.

Here is an example of this defect, before and after correction in the wm volume:

The slice shown is 128. You would of course scroll through coronally and edit any slices where this defect was present. See the next page for what this subject's inflated surface looks like after all defects (including the temporal lobe) are corrected and the surface was recreated.

Arrow indicates corrected temporal lobe defect on inflated surface, medial aspect:

1) After saving the volume in TkMedit’s File menu, go to csurf’s ‘Subject Tools’ menu, and select ‘Create Surface’, then choose the appropriate hemisphere(s) to process:

2) Select ‘Create Surface’ in the box, where you can then monitor its progress:

3) An ALERT box will inform you when the ‘Create Surface’ process is complete:

Topology Fixing Overview

Once the surface is free of large anatomically derived defects, the automated topology fixer can be used to remove smaller topological defects. The Fix Surface Topology process takes approximately 2 hrs per hemisphere to run to completion.

This process can be selected under the SubjectTools menu:

Here is an outline of the subprocesses of Fix Surface Topology, and the respective files they generate:

Part 1: quick sphere RH surface

(inflates the topologically defective right hemisphere surface to a sphere)

The output file written by this procedure is:

surface: $SUBJECTS_DIR/$name/surf/rh.qsphere

Part 2: quick sphere LH surface

(inflates the topologically defective left hemisphere surface to a sphere)

The output file written by this procedure is:

surface: $SUBJECTS_DIR/$name/surf/lh.qsphere

Part 3: fix topology RH surface

(automatic topology fixing of the right hemisphere)

The output files written by this procedure are:

surface: $SUBJECTS_DIR/$name/surf/rh.orig

curv: $SUBJECTS_DIR/$name/surf/rh.defect_status

curv: $SUBJECTS_DIR/$name/surf/rh.defect_labels

Part 4: fix topology LH surface

(automatic topology fixing of the left hemisphere)

The output files written by this procedure are:

surface: $SUBJECTS_DIR/$name/surf/lh.orig

curv: $SUBJECTS_DIR/$name/surf/lh.dect_status

curv: $SUBJECTS_DIR/$name/surf/l.defect_labels

Part 5: resmooth RH white matter

(smooths rh.orig surface)

The output files written by this procedure are:

surface: $SUBJECTS_DIR/$name/surf/rh.smoothwm

curv: $SUBJECTS_DIR/$name/surf/rh.curv

curv: $SUBJECTS_DIR/$name/surf/rh.sulc

curv: $SUBJECTS_DIR/$name/surf/rh.area

Part 6: resmooth LH white matter

(smooths lh.orig surface)

The output files written by this procedure are:

surface: $SUBJECTS_DIR/$name/surf/lh.smoothwm

curv: $SUBJECTS_DIR/$name/surf/lh.curv

curv: $SUBJECTS_DIR/$name/surf/lh.sulc

curv: $SUBJECTS_DIR/$name/surf/lh.area

Part 7: reinflate RH white matter

(inflates rh.orig surface)

The output file written by this procedure is:

surface: $SUBJECTS_DIR/$name/surf/rh.inflated

Part 8: reinflate LH white matter

(inflates lh.orig surface)

The output file written by this procedure is:

surface: $SUBJECTS_DIR/$name/surf/lh.inflated

Fix Surface Topology Guide

Once the manual defects have been corrected, the smaller topological defects need to be fixed. These smaller, often invisible defects can be corrected using an automated procedure, Fix Surface Topology, which takes about 4 hours to run for both hemisphers.

In the ‘Subject Tools’ menu of csurf, select ‘Fix Surface Topology’, then choose ‘BOTH’.

Click on ‘Fix Topology’:

You can monitor the progress of the steps involved:

A dialog box will alert you when ‘Fix Topology’ is complete.

Visual Comparison of Inflated Surfaces (lateral aspect)

first inflated surface showing defects à à after editing defects and reinflation à

Make Final Surfaces Overview

This automated step starts a two-part background process to create the final left and right hemisphere cortical surfaces. The gray/white boundary is called ?h.white and the gray/csf boundary is called ?h.pial. The white and pial surfaces are used to estimate the cortical thickness at all locations on the cortical surface. The thickness estimates are stored in a curv file called ?h.thickness. Make Final Surface should only be run once the surface editing is complete (i.e. the surface is topologically correct). The two-part processes take a total of approximately 6 hrs per hemisphere to run to completion, after which you can view the 2-D surfaces it outputs (rh.pial, lh.pial, rh.white and lh.white) in surfer, or in TkMedit as 3-D overlays (use TkMedit’s ‘Display’ menu to display them, once loaded in).

Part 1: Make Final Right Hemisphere Surfaces

The output files written by this procedure are:

surface: $SUBJECTS_DIR/$name/surf/rh.white

surface: $SUBJECTS_DIR/$name/surf/rh.pial

curv: $SUBJECTS_DIR/$name/surf/rh.thickness

Part 2: Make Final Left Hemisphere Surfaces

The output files written by this procedure are:

surface: $SUBJECTS_DIR/$name/surf/lh.white

surface: $SUBJECTS_DIR/$name/surf/lh.pial

curv: $SUBJECTS_DIR/$name/surf/lh.thickness

This is the next step after the topologically correct surfaces are available.

To start this final step in the reconstruction, go to csurf’s `Subject Tools’ menu and select ‘Make Final Surfaces’. Choose ‘both’ hemispheres when prompted.

A dialog box will alert you when the process is complete.

Make Final Surface generates the two additional surfaces shown here:

1. The white surface:



2. The pial surface:

Make Final Surface Guide

The Make Final Surface procedure allows for the generation of the final white and gray surfaces. These surfaces are useful since they provide representations of the white and the gray surfaces, so that measurements such as cortical thickness and surface area can be computed. To start this final step in the reconstruction, go to csurf’s ‘Subject Tools’ menu and select ‘Make Final Surfaces’. Choose ‘both’ hemispheres when prompted.

Now click on the ‘MakeFinalSurface’ button:

A dialog box will alert you when the process is complete.