3D Volume Data Fusion
A High-level Introduction
To learn more, see selections
under 3D Volume Fusion
at left.
This presentation and images were originally created by Bill Budenholzer
for RAHD Oncology Products. It has since been modified.
|
Shown is one of the workspace layouts, with orthogonal views of a
PET on
the left, and the fused PET and CT blended on the right.
|
A Simple Walkthrough
|
 |
Start with a DICOM transfer, for example, of the axial slices from a PET image set
showing the disease. |
| Or, select MR, CT or multiple studies to be merged into one coordinate
space. |
 |
| Next, select the CT image set that
represents the reference volume, defining the image coordinate set that you want to use.
|
|
Any known image format can be fused with the reference image set. The only
limitations are those encountered in the clinical environment. |
|
| Identify five or more complimentary anatomical landmarks in each image set.
This nominally takes from 5 to 15 minutes (once familiar with the process),
depending on the modalities and anatomical regions. Various tools are provided to aid in landmark selection. |
 |
 |
A complete array of tools exist to assist with viewing the image and
selecting appropriate anatomy. Shown are a skeletal isosurface and landmarks in their 3D positions, displayed in an interactive window. |
| Select the fusion algorithm that will best merge the
target into the reference 3-D volume data set.
The images will be quickly and accurately fused, based on the
landmarks and the chosen algorithm.
|
 |
|
Choice of Algorithms
Rigid body -- Best-fit to landmarks via rigid body transform,
limited to rotation, translation, and uniform scaling.
Affine -- Best-fit to landmarks via rigid body
transform with stretching or skewing. First order warp -- Best-fit to landmarks using the 1st order terms of a 2nd order polynomial.
Works well for locally constrained warping of soft tissue anatomy such as
lung or abdomen.
Second order warp -- Best-fit to landmarks using the
full 2nd order polynomial to make the distortions necessary to fit large
disparities in anatomy.
Manual - Interactive user settings of rotation, translation, and scale,
in each orthogonal dimension; for example, to 'tweak' the fusion results.
|
|
|
An array of 3D and 2D tools are provided for analysis, landmark selection, and validation of results.
|
|
| Overlay and Crossfade
 |
Tools such as image overlay, side by side comparisons,
isoline overlay, and statistical assessment are yours to assist with viewing the image
volumes, selecting appropriate anatomy, and analyzing the fusion results. Shown is a SPECT on the left, and SPECT/CT fusion on the right. |
|
Any Level - Any Cut
|
|
Isolines
 |
Library of Color Map Tables
 |
|
Validation of Fusion Results Quantitative measurement -- of individual landmark errors.
Qualititative visual analysis -- with various 3D and 2D tools designed specifically for this purpose.
A complete image analysis toolkit assures accurate results. You know you've got it right.
|
|
Move your mouse over the pictures below to see the difference.
Mouse-over image to see difference |
Here we see image sets obtained with the
patient in very different positions. A supine MR on the left. A prone CT on the right, with greater than 20 degree head tilt.
Seemingly incompatible orientations are easily fused.
Mouse-over the image to see the fused MR on top of the original CT. |
| This fusion accurately corrected the differences in the
supine MR and prone CT with greather than 20 degree tilt of the head.
Mouse-over the image to see the fused MR overlaid in blue on top of the original CT. |
Mouse-over image to see difference |
Mouse-over image to see difference |
Isolines generated on the Reference image set
can be projected onto the Fused image set, or vice versa, to help verify the quality of the fused results.
Mouse-over image to see the isolines. |
|
Tools are provided to visualize the Reference and Fused images superimposed, or overlaid, to facilitate the visual analysis.
Each can be set to a different color map.
Mouse-over the images to see the overlay tool.
Blend -- Variable blending of the Reference and Fused images.
Viewport -- A rectangular viewport can be sized and moved across the image, looking through
one image onto the other.
Split Screen -- A movable vertical or horizontal split screen allows
quick assessment across the entire image.
|
Mouse-over images to see difference |
Quantitative analysis of fusion results
showing individual landmark errors (in mm), and cumulative error statistics. |
|
Landmark List - Warped Image
Warped-Reference Landmark
Distance Differences (mm)
- 3D Distance From Reference Landmark 0.53
- 3D Distance From Reference Landmark 1.94
- 3D Distance From Reference Landmark 1.44
- 3D Distance From Reference Landmark 2.17
- 3D Distance From Reference Landmark 2.54
- 3D Distance From Reference Landmark 1.71
- 3D Distance From Reference Landmark 1.24
- 3D Distance From Reference Landmark 1.88
- 3D Distance From Reference Landmark 2.15
Cumulated Statistics
Standard Deviation 0.60, Standard Error of the Mean 0.20
Mean Value 1.73, Maximum Error 2.54, Minimum Error 0.53
|
|
To learn more, see selections
under 3D Volume Fusion
at left.
Back to Top
|
|
