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.
High performance computing for deformable image registration: Towards a new paradigm in adaptive radiotherapy
Rent:
Rent this article for
USD
10.1118/1.2948318
/content/aapm/journal/medphys/35/8/10.1118/1.2948318
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/35/8/10.1118/1.2948318

Figures

Image of FIG. 1.
FIG. 1.

Deformable image registration in lung imaging. Representative 2D imaging planes presented here are based on volumetric deformable image registration using demon’s method: (a) reference image, (b) target image, (c) deformed image using demons for deformable image registration (DIR) on imaging (a). (d) Difference image of the reference image and target image (i.e., prior to DIR). (e) Difference image of the reference image and deformed image (i.e., following DIR).

Image of FIG. 2.
FIG. 2.

Auto-recontouring for lung imaging using DIR. Physician drawn contours (drawn on separate reference FB CT) and computer generated contours, using DIR based mapping, have been overlaid on a target CT (i.e., phase 5). The LUNG-RT and GTV-fb (brown and green colored lines), representing outlining the right lung and gross tumor, respectively, are physician drawn contours on the reference CT. The auto LUNG-RT and auto GTV-fb (red and blue colored lines) contours are the corresponding computer generated contours on target CT based on DIR.

Image of FIG. 3.
FIG. 3.

Performance comparisons between the CPU (single thread and dual thread) and GPU based implementations of DIR as a function of image size. GPU implementation is times faster than the corresponding single thread CPU implementation, and times faster than the corresponding dual thread CPU implementation. All three implementations exhibit a linear dependence on image size.

Image of FIG. 4.
FIG. 4.

Computational efficiency of GPU programming for DIR. The effect on data size on GPU computational efficiency was explored. Here we superimposed our GPU (NVIDIA 8800 GTX using CUDA programming language) results, alongside that of Sharp et al. who used NVIDIA 8800 GTS with Brook programming language.

Tables

Generic image for table
TABLE I.

Maximum, mean and standard deviation of the difference in correlation coefficients between the CPU implementations and the GPU implementation in 100 iterations

Generic image for table
TABLE II.

Performance comparison between the CPU and GPU based implementations of DIR for clinical imaging data.

Generic image for table
TABLE III.

Performance comparison between the CPU ( Intel Dual Core with RAM) and GPU (NVidia 8800 GTX) based implementations of DIR (i.e., demons) for clinical imaging data for Sharp et al.’s GPU implementation (see Ref. 28).

Loading

Article metrics loading...

/content/aapm/journal/medphys/35/8/10.1118/1.2948318
2008-07-11
2014-04-25
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: High performance computing for deformable image registration: Towards a new paradigm in adaptive radiotherapy
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/35/8/10.1118/1.2948318
10.1118/1.2948318
SEARCH_EXPAND_ITEM