|
| |
Registration Dialog
  
Use this dialog to register two
sets of images, align a single set, or warp one set to another using contours. Registration
and warping can be performed together on 2 sets from different modalities (reference).
Note
This dialog uses the
Automated Image Registration
package (alignlinear) from UCLA. You must download and install the
AIR software for the AIR method of registration o function.
The programs must also be in your search path for this method to work. To
check, type "which alignlinear" on a command-line terminal. If you do not
get the full path to the program, you need to add the path to the programs in
your .cshrc or .profile. See your System Administrator for help.
Most of the settings for the Registration Dialog can be set through
a resource file
that is loaded at the start of the program.
 |
Example Registering using Head in Hat Method
Warning
The registration process assumes that the 2 sequences being registered have
the same slice thickness. If they do not,
click here for
instructions on how to modify the slice thickness of a data set to match
another data set.
Load a central location, for all scanning sequences, from a
patient study and use the left button in the Browser Window (see
Selecting Images to Display) to check alignment between the scanning sequences. Make a note of which scanning
sequences are misaligned and use the following instructions to align.
Note
Delete the browser before
continuing with Registration.
Use Split Display in the Display Windows to speed processing
during registration.
During registration, ensure that both sequences being registered
contain as much of the patient as possible, including the top of the skull
without brain.
-
Load all locations of one
scanning sequence from a patient study. This will be referenced as the base set
into Browser #1.
-
Load all locations of another
scanning sequence from the same patient study into Browser #2. This
sequence will be registered to the images loaded into Browser #1.
-
Link the browsers for the two sets.
Note
If the contrast between the two sets of images are
substantially different, do not link the browsers. You will need to
repeat steps 4 - 6 for the second browser with different Hi values.
-
Select
Contour in the Region of Interest/Volume of Interest
Dialog.
-
Enter 0 for the
Lo value. To select the Hi value: move cursor over the skin area of the
image, observe the initial peak in gray level as the skin area is entered
(lower-left of the Display Window). Choose a Hi value slightly less than
this peak (~75% of the value).
-
Use the left
button to select an area outside the skin, the upper-right quadrant of the image
works best. This will display the image contour. Ensure the contour
is smooth and free from gross discontinuities, small breaks in the contour and
outlines of the nose and ears will not effect the outcome. Adjust the Hi
value as necessary or erase parts of the contour using
Erase Irregular and
redraw manually using Irregular.
Note
Choosing a different point on the image may drastically change the contour
outline.
Ensure, the Region of Interest Dialog displays the Browser # of the images you
are working on. If necessary, move mouse over desired browser and continue
working. This will reset the Browser # of the Region of Interest Dialog.
-
Press
Start Macro, select a point outside the skin, press
End Macro, and press
Run Macro.
-
Use the
left button in the Browser Window
to check all locations for gross errors between the two sets.
- If any gross errors exist, press another point in the
incorrect image to repair. If this does not repair the contour enter
a different value for the Hi value and press another point in the incorrect image.
(Step 4)
Note
If you did not link the browsers in Step 3, Repeat steps
4-6 for the second browser with different Hi values.
-
Use the
right button in the Browser Window
to delete any images that do not have sufficient anatomy.
-
Select
Analyze->Registration->Register
from the Main Window.
-
Press
Match. The residual value will be displayed in
the information window. Press Match several more times until the
residual value reaches a minimum value and then begins to increase..
Note
The residual value can vary significantly between
patients and even between different sets of the same patient. Do not
wait for the residual value to fall below specific level.
-
Upon the
first increase in the residual value, press Backup.
-
Press
Reslice.
-
Enable the two
browsers for Link Displays
in the Main Window.
-
Use the
left button in the Browser Window to check
registration between the two sets, checking ventricles and skull.
-
Set
and Send Link as a
in the Main Window to Region. This will copy the region of interest
created in steps b and c to be displayed on both sets of images.
-
Use
Threshold or
Slice to draw regions.
-
Use
Edgi to get outline.
If misregistered,
delete the browser with the resliced
images and repeat beginning at Step 4.
You may need to select different anatomy, other than skin, for the contour or the point used to start
the contour may need to be changed.
If registration is successful,
delete the browsers for
the scanning sequence that was resliced. Start at Step 2 if there are more
scanning sequences to register.
|
|
Example
Registering using the AIR Method
Warning
The registration process assumes that the 2 sequences being registered have
the same slice thickness. If they do not,
click here for
instructions on how to modify the slice thickness of a data set to match
another data set.
Note
This method uses programs provided by
Automated Image Registration
package (alignlinear) from UCLA. Load a central location, for all scanning sequences, from a
patient study and use the left button in the Browser Window (see
Selecting Images to Display) to check alignment between the scanning sequences. Make a note of which scanning
sequences are misaligned and use the following instructions to align.
-
Load all locations of one
scanning sequence from a patient study. This will be referenced as the base set
into Browser #1.
-
Load all locations of another
scanning sequence from the same patient study into Browser #2. This
sequence will be registered to the images loaded into Browser #1.
-
In both
browsers, select a central location.
-
Using the mouse,
get an average value for white matter from each set and calculate a
scaling factor = (white matter value browser #1) / (white
matter value browser#2).
Note
If the scaling factor is around 1, you can skip steps
5 and 6.
-
Select
Roi/Voi->Create/Modify from the Main
Menu.
-
Select
All Images (2D).
-
Select
Automatic.
-
Select
Fill.
-
Use left
button to select several white matter areas in Browser # 2, then
press the right button to get a white matter region encompassing
most of the white matter.
-
Select
Roi/Voi Processing.
-
Press
FilG (make sure the
Region of Interest/Volume of Interest Dialog has Browser 2 in the title
bar, you can pass the mouse over Browser #2 to change it).
Note
Check the results by looking at several locations through
the image set. If not satisfactory, clear the
values for FilG
and repeat steps 5 b and c.
-
Select
Current Image.
-
Press
Cancel to close the Region of Interest/Volume of Interest Dialog.
-
Select
Analyze->Miscellaneous->Math.
-
Enter the
scaling factor calculated previously in the
1st Value field.
-
Press
Muk.
-
Select
Browser #2 for Set 1(it should be selected already), with Browser
#3 as the Output Browser (it should already be selected, as
well).
-
Press OK.
-
Press
Cancel to close the Math Dialog.
-
Select
Analyze->Registration->Register
from the Main Window.
Note
If you did not perform steps 5 and 6 (normalization) use
Browser #2, not Browser #3, for the Reslice Set.
-
Select
AIR for the Match Method. See the
note at the top for important information.
-
Using the
left button over the display image, determine a threshold to
include the brain and skin, but not the background for Browser #1 and
Browser #3.
Note
You can use
Threshold in the Region of Interest/Volume of Interest dialog to
check the values.
-
Enter the
threshold for Browser #1 in -t1 and the
threshold for Browser #3 in -t2.
-
Press
Match.
-
Select
Browser #1 for the Base Set and Browser #3 for the Reslice Set.
-
Press OK.
-
Press
Reslice.
-
Select
Browser #1 for the Base Set and Browser #2 for the Reslice Set
-
Press OK.
-
Link Browsers 1,
2, and 4 (set and Send Link as a to Region).
Note
If you skipped steps 5 and 6, link Browsers 1, 2, and 3.
-
Select
Automatic, Fill.
-
Press the
left button over white matter in Browser #1, then press the right
button several times to completely cover white matter, but not any
other anatomy.
-
Select
Morphological.
-
Press
Edgi.
|
|
Example
Registering using the Mutual Information Method
Warning
The registration process assumes that the 2 sequences being registered have
the same slice thickness. If they do not,
click here for
instructions on how to modify the slice thickness of a data set to match
another data set.
Load a central location, for all scanning sequences, from a
patient study and use the left button in the Browser Window (see
Selecting Images to Display) to check alignment between the scanning sequences. Make a note of which scanning
sequences are misaligned and use the following instructions to align.
-
Load all locations of one
scanning sequence from a patient study. This will be referenced as the base set
into Browser #1.
-
Load all locations of another
scanning sequence from the same patient study into Browser #2. This
sequence will be registered to the images loaded into Browser #1.
-
In both
browsers, select a central location.
-
Select
Analyze->Registration->Register
from the Main Window.
-
Select
Mutual Information for the Match Method.
-
Using the
left button over the display image, determine a range of values to
cover the normal brain anatomy for Browser #1 and
Browser #2.
Note
You can use Slice in the Region of Interest/Volume of Interest dialog to
check the values.
-
Enter the
slice values for Browser #1 in Base and the
slice values for Browser #2 in Reslice.
-
Press
Match.
-
Select
Browser #1 for the Base Set and Browser #2 for the Reslice Set.
-
Press OK.
-
Press
Reslice.
-
Select
Browser #1 for the Base Set and Browser #2 for the Reslice Set
-
Press OK.
-
Link Browsers 1,
2, and 3 (set and Send Link as a to Region).
-
Select
Automatic, Fill.
-
Press the
left button over white matter in Browser #1, then press the right
button several times to completely cover white matter, but not any
other anatomy.
-
Select
Morphological.
-
Press
Edgi.
|
|
Example Aligning One Image Set
-
Load all locations of one
scanning sequence from a patient study.
Select
Contour in the Region of Interest/Volume of Interest
Dialog.
-
Enter 0 for the
Lo value. To select the Hi value: move cursor over the skin area of the
image, observe the initial peak in gray level as the skin area is entered
(lower-left of the Display Window). Choose a Hi value slightly less than
this peak (~75% of the value).
-
Use the left
button to select an area outside the skin, the upper-right quadrant of the image
works best. This will display the image contour. Ensure the contour
is smooth and free from gross discontinuities, small breaks in the contour and
outlines of the nose and ears will not effect the outcome. Adjust the Hi
value as necessary or erase parts of the contour using
Erase Irregular and
redraw manually using Irregular.
Note
Choosing a different point on the image may drastically change the contour
outline.
Ensure, the Region of Interest Dialog displays the Browser # of the images you
are working on. If necessary, move mouse over desired browser and continue
working. This will reset the Browser # of the Region of Interest Dialog.
-
Press
Start Macro, select a point outside the skin, press
End Macro, and press
Run Macro.
-
Use the
left button in the Browser Window
to check all locations for gross errors in the data set.
- If any gross errors exist, press another point in the
incorrect image to repair. If this does not repair the contour enter
a different value for the Hi value and press another point in the incorrect image.
(Step 1)
-
Use the
right button in the Browser Window
to delete any images that do not have sufficient anatomy.
-
Select
Rotate if you want to perform rotation when
aligning. Deselect it to translate only.
-
Press
Align, each image in turn, will be translated and
rotated (if selected).
|
|
Example Warping One Image Set to
An Expected Image Set
-
Load all locations of one scanning sequence from a patient
study. This will be referenced as the expected set into Browser #1.
-
Load all locations of another scanning sequence from the
same patient study into Browser #2. This sequence will be warped to the
images loaded into Browser #1.
-
Select 3D
in Warp Dimensions.
-
Select
Contour in the Region of Interest/Volume of Interest
Dialog.
-
Enter 0 for the
Lo value. To select the Hi value: move cursor over the skin area of the
image, observe the initial peak in gray level as the skin area is entered
(lower-left of the Display Window). Choose a Hi value slightly less than
this peak (~75% of the value).
-
Use the left
button to select an area outside the skin, the upper-right quadrant of the image
works best. This will display the image contour. Ensure the contour
is smooth and free from gross discontinuities, small breaks in the contour and
outlines of the nose and ears will not effect the outcome. Adjust the Hi
value as necessary or erase parts of the contour using
Erase Irregular and
redraw manually using Irregular.
Note
Choosing a different point on the image may drastically change the contour
outline.
Ensure, the Region of Interest Dialog displays the Browser # of the images you
are working on. If necessary, move mouse over desired browser and continue
working. This will reset the Browser # of the Region of Interest Dialog.
-
Press
Start Macro, select a point outside the skin, press
End Macro, and press
Run Macro.
-
Use the
left button in the Browser Window
to check all locations for gross errors between the two sets.
- If any gross errors exist, press another point in the
incorrect image to repair. If this does not repair the contour enter
a different value for the Hi value and press another point in the incorrect image.
(Step 3)
-
Use the
right button in the Browser Window
to delete any images that do not have sufficient anatomy.
-
Repeat steps 4-7
for the warp images.
-
Press
Exp Pts for the expected images (Browser #1)
and Wrp Pts for the images to warp (Browser
#2).
-
Repeats Steps
6-8 for as many anatomical features you would like to use for calculating the
correction table. You can use any region
of interest functions to create the outlines of the anatomical features, as
long as you have the outline only (Edgi
will create an outline of any solid region of interest).
Important
The image border should also be used for one of the features to prevent
unexpected results outside the area of interest. The
Invert and
Edgi functions in the Region of
Interest dialog can be used to create this.
Note
You can start over by pressing Clr Exp and
Clr Wrp.
You can display your current expected and warped points by Pressing
Dis Exp and Dis
Wrp.
-
Select
Analyze->Registration->Register from the Main
Window.
-
Press
Warp Table. The contours for each set will be saved with "exp"
for Browser #1 and "warp" for Browser #2. Vectors will be drawn on images
displaying the direction of warp for each pair of warp -> expected points.
These vectors will also be saved for each image with "warp_to_exp" as the
extension.
-
Press
Correct.
|
|
Dialog Items
 |
Match
Press Match
to determine the translation and rotation parameters to register one set of
images to another base set of images.
 |
Select
Head/Hat
to calculate the translation and rotation parameters using a Head in Hat
method. A contour should
be drawn of the same anatomical feature for each set, but does not need to
be on all images.
Note
Both data sets must be the same image size and thickness. The x and y
pixel sizes can be different.
Load all images from both sets, although only draw the contours where the
anatomical feature exists.
A residual value will be printed in the
information window indicating the error between the two sets, the lower
the value the better. Press Match repeatedly to improve the
residual value, stop when the value increases.
Backup can be pressed to use the prior match with the lower residual
value.
|
|
Select
AIR
to calculate the translation and rotation parameters using the
Automated Image Registration
package (alignlinear) from UCLA. See the note
at the top for important information.
|
Enter the threshold value for the Base Set in -t1.
|
|
Enter the threshold value for the Reslice Set in -t2.
|
|
Enter the
command-line options, including model number, for alignlinear in
Options.
-m
model_menu_number [options]
Model Menu:
3-D models:
6. rigid body 6 parameter model
7. global rescale 7 parameter model
9. traditional 9 parameter model (std must be on AC-PC
line)
12. affine 12 parameter model
15. perspective 15 parameter model
2-D models (constrained to in-plane, no interpolation):
23. 2-D rigid body 3 parameter model
24. 2-D global rescale 4 parameter model
25. 2-D affine/fixed determinant 5 parameter model
26. 2-D affine 6 parameter model
28. 2-D perspective 8 parameter model
options:
(defaults in parentheses)
[-b1 FWHM_x(0.000000e+00) FWHM_y(0.000000e+00)
FWHM_z(0.000000e+00)] (standard file)
[-b2 FWHM_x(0.000000e+00) FWHM_y(0.000000e+00)
FWHM_z(0.000000e+00)] (reslice file)
[-c convergence_threshold(0.000010)]
[-d] use static partitioning (like AIR3.0 and earlier)
[-e1 standard_file_mask]
[-e2 reslice_file_mask]
[-f initialization_file]
[-fs scaling_initialization_file]
[-g termination_file [overwrite?(y/n)]]
[-gs scaling_termination_file [overwrite?(y/n)]
[-h halt_after_(5)_iterations_without_improvement]
[-j] use non-positive definite Hessian matrices (not
validated, use at your own risk)
[-p1 segment_standard_file_into_(1)_partitions]
[-p2 segment_reslice_file_into_(1)_partitions]
[-q] assume non-interaction of spatial parameter
derivatives
[-r repeated_iterations(25)]
[-s initial_sampling(81) final_sampling(1)
sampling_decrement_ratio(3)]
[-t1 threshold_standard_file(75)]
[-t2 threshold_reslice_file(75)]
[-v] verbose reporting of interim results
[-x cost_function(1)]
1. standard deviation of ratio image
2. least squares
3. least squares with intensity rescaling
[-z] (no additional argument; turns on pre-alignment
interpolation) |
|
|
Select
Mutual
Information to calculate the rotation and translation parameters by
minimizing the entropy of the intensity plot. (reference)
Note
This option is currently under development and is not available outside
the Henry Ford Health System.
|
|
|
Backup
Press Backup
to undo the last Match operation (Head/Hat
only)
|
|
Reslice
Press Reslice
to translate and rotate the second image data set to match the first (base)
image data set, after performing the Match function.
Note
The same match results can be used on multiple browsers by pressing Reslice
again and selecting the other browser for the Reslice set. This
information is not saved once the user quits the program.
|
|
Align
Press Align
to align a set of images using contours of an anatomical feature. The
center of mass of each contour is used to translate each image. The
error between each contour is used to determine rotation of each image
(first by 10 degrees, 5 degrees, and 1 degree).
|
Rotate option
Select Rotate
to perform rotation when aligning. If not selected, only translation
will be performed.
|
|
|
Warp Dimensions
|
Select 2D
for single slice processing of warp procedures
|
|
Select 3D
for volume processing of warp procedures.
|
|
|
Expected Points
Press Exp Pts
to sample the contour radially around the center and add to the expected
points list.
|
Select Rays
(2d) to sample warp and expected points by rotating a ray clockwise
starting at the top mid-line.
|
|
Select Rays
(3d) to sample warp and expected points by rotating a ray clockwise
starting at the top-midline, for each location.
|
|
Select Edge
(2d) to sample warp and expected points by following the contour,
starting at the top-midline, and sampling an equal number of points.
|
|
|
Warp Points
Press Warp
Pts to sample the contour radially around the center and add to the warp
points list.
|
|
Clear Expected
Press Clr Exp
to clear the expected points list.
|
|
Clear Warp
Press Clr Wrp
to clear the warp points list.
|
|
Display Expected
Press Dis Exp
to display the current contents of the expected points list as a region of
interest on each image. |
|
Display Warp
Press Dis Wrp
to display the current contents of the warp points list as a region of
interest on each image. |
|
Warp Table
Press Warp
Table
to calculate the warp correction tables for the current set of warp and
expected points. The correction is displayed graphically as vector
regions of interest on each image connecting the warp point to its
corresponding expected point. The sampled warp and expected points and
the correspondence region are automatically saved as region of interest
files.
|
Select
Bookstein's to calculate the correction tables using
Bookstein's method.
This is a look-back method.
|
|
Select
Look-Back to use bi-linear interpolation of the mapping of expected
points to warp points to calculate the correction tables.
|
|
Select
Look-Forward to use bi-linear interpolation of the mapping of warp
points to expected points to calculate the correction tables. |
|
|
Correct
Press Correct
to correct the selected images using the warp correction tables that were calculated
or create images of the X, Y, or Z component of the warp correction tables.
|
Select
Look-Back to correct the selected images using a correction table created
with a look-back method.
|
|
Select
Look-Forward to correct the selected images using a correction table
created with a look-forward method.
|
|
Select Delta-X
to display the X component of the correction table as images.
|
|
Select Delta-Y
to display the Y component of the correction table as images.
|
|
Select Delta-Z
to display the Z component of the correction table as images.
|
|
|
Cancel
Press Cancel
to remove the Registration dialog. |
|
Help
Press Help
to display help for the Registration dialog. |
|
Back to
Eigentool
|
