1887
banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
Geometric feature-based multimodal image registration of contrast-enhanced cardiac CT with gated myocardial perfusion SPECT
Rent:
Rent this article for
USD
10.1118/1.3253301
/content/aapm/journal/medphys/36/12/10.1118/1.3253301
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/36/12/10.1118/1.3253301

Figures

Image of FIG. 1.
FIG. 1.

Examples of CT (top) and MPS (bottom) volume cross sections illustrating image variations between patients: (a) and (b) show coronal views for two different patients. (c) and (d) show the axial/transverse views for the respective patients.

Image of FIG. 2.
FIG. 2.

An example of myocardial segmentation of MPS: (a) shows transverse views for an abnormal case (top) and a normal case (bottom); (b) shows respective contours from segmentation overlaid on the images from (a); (c) illustrates the blood pool , myocardial , and extracardiac regions; (d) shows 3D contours obtained from the myocardial segmentation, endocardium (surface), and epicardium (mesh). Note the robust segmentation result despite abnormality in MPS.

Image of FIG. 3.
FIG. 3.

Overview of the registration procedure.

Image of FIG. 4.
FIG. 4.

An example of finding the best matching phase of gated MPS by minimizing the registration cost functional: (a) plot of cost functional value across different phases of MPS; (b) overlaid registration result of phase 9 MPS on CT perfusion; (c) overlaid registration result of best matching phase (16) with CT perfusion. Note that the best matching MPS phase is visually well correlated with CT.

Image of FIG. 5.
FIG. 5.

“Motion-frozen” technique: All MPS frames are registered to a common reference frame , which is the best matching MPS phase for the CT volume. TPS-based warping is used to deform and map each frame to . The control points from the segmentation contours serve as landmark points for TPS warping and corresponding points from the source and target contours determine the mapping. Final motion-frozen image is obtained by averaging over all the warped volumes. Note that perfusion defects are clearly seen in the final image (final column).

Image of FIG. 6.
FIG. 6.

An example of registration results: original CT image (top), images before (middle), and after (bottom) registration are shown. Errors were 2, 1, and for translation and 1°, 1°, and 0° for rotation, as compared to visual alignment.

Image of FIG. 7.
FIG. 7.

An example of registration results (CT perfusion): original CT image (top), images before (middle), and after (bottom) registration are shown. Errors were 1, 0, and for translation and 5°, 2°, and 3° for the rotation, as compared to visual alignment.

Image of FIG. 8.
FIG. 8.

Comparison of NMI based method and proposed method: (a) registration result using NMI based method; (b) registration result using the proposed method; (c) the NMI values for translational offsets from the ground truth for each axis; (d) the cost values of proposed method for translational offsets from the ground truth for each axis. Note that the anatomical features are very different and the image segmentation will aid the registration of MPS to CT. Additionally, cost value of the proposed method is optimal at the origin, whereas NMI does not provide any such robust optimum.

Image of FIG. 9.
FIG. 9.

Registration errors: (a) translational errors of stress/rest MPS and CT obtained from site A; (b) rotational errors of stress/rest MPS, and CTA obtained from site A; (c) translational errors of stress/rest MPS, and CT obtained from site B; (d) rotational errors of stress/rest MPS, and CT obtained from site B; (e) translational errors of stress/rest MPS and perfusion CT; (f) rotational errors of stress/rest MPS and perfusion CT.

Image of FIG. 10.
FIG. 10.

Capture ranges of the proposed method: (a) translation errors for different misalignment initializations along axis; (b) translation errors for different misalignment initializations along axis; (c) translation errors for different misalignment initializations along axis.

Image of FIG. 11.
FIG. 11.

Capture ranges of the NMI based method: (a) translation errors for different misalignment initializations along axis; (b) translation errors for different misalignment initializations along axis; (c) translation errors for different misalignment initializations along axis.

Tables

Generic image for table
TABLE I.

Registration error and interobserver variability.

Generic image for table
TABLE II.

Evaluation of best matching phase.

Loading

Article metrics loading...

/content/aapm/journal/medphys/36/12/10.1118/1.3253301
2009-11-05
2014-04-18
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: Geometric feature-based multimodal image registration of contrast-enhanced cardiac CT with gated myocardial perfusion SPECT
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/36/12/10.1118/1.3253301
10.1118/1.3253301
SEARCH_EXPAND_ITEM