Logo image
Back
Chapter 47 - Deformable Volumetric Registration Using B-Splines
Book chapter

Chapter 47 - Deformable Volumetric Registration Using B-Splines

James Shackelford, Nagarajan Kandasamy and Gregory Sharp
GPU Computing Gems Emerald Edition, pp 751-770
2011
DOI: https://doi.org/10.1016/B978-0-12-384988-5.00047-4

Abstract

This chapter shows how to develop a B-spline-based deformable registration algorithm within the single instruction multiple thread (SIMT) model to effectively leverage the large number of processing cores available in modern GPUs. The focus is on improving processing speed without sacrificing quality to make the use of deformable registration more viable within the medical community. A registration is called rigid if the motion or change is limited to global rotations and translations, and is called deformable when the registration includes complex local variations. Rigid matching is appropriate for serial imaging of the skull, brain, or other rigidly immobilized sites. Deformable registration is appropriate for almost all other scenarios and is useful for many applications within medical research, medical diagnosis, and interventional treatments. To date, several deformable registration algorithms have been proposed and validated, including demons registration, viscous fluid registration, B-spline registration, and thin-plate splines. Performance results on a Tesla C1060 show that for large images, the GPU algorithm achieves a speedup of 15 times over a single-threaded CPU implementation. Moreover, the CPU and GPU versions achieve near-identical registration quality. This chapter introduced a grid-alignment technique to greatly reduce the complexity of B-spline-based registration. The main ideas underlying the aligned-grid method have been used to develop a highly parallel design for the GPU.

Metrics

36 Record Views

Details

Logo image