Efficient GPU-Accelerated Elastic Image Registration
X-Ray Predevelopment, Philips Medical Systems, Best, the Netherlands
Proceedings of the Sixth IASTED International Conference on Biomedical Engineering, BioMED ’08
@conference{ruijters2008efficient,
title={Efficient GPU-accelerated elastic image registration},
author={Ruijters, D. and ter Haar-Romeny, B.M. and Suetens, P.},
booktitle={Proceedings of the Sixth IASTED International Conference on Biomedical Engineering},
pages={419–424},
year={2008},
organization={Acta Press}
}
Elastic intra-patient registration can be used to correct for local motion within biomedical images. The application of elastic registration during interventional treatment is seriously hampered by its considerable computation time. The Graphics Processing Units (GPU) can be used to accelerate the calculation of such elastic registrations, without changing the basic registration algorithm. This article discusses how elastic image registration, using cubic B-spline based deformation fields, efficiently can be approached to make use of the vast processing power of the GPU. Our approach employs an efficient GPU-based cubic B-spline deformation field, and also calculates the similarity measure and its derivative on the GPU.
February 26, 2011 by hgpu
