Parallel and memory efficient multimodal image registration for radiotherapy using normalized gradient fields

We introduce a new highly parallel and memory efficient deformable image registration algorithm to handle challenging clinical applications. The algorithm is based on the normalized gradient fields (NGF) distance measure and Gauss-Newton numerical optimization. By carefully analyzing the mathematical structure of the problem, a matrix-free Hessian-vector multiplication for NGF is derived, giving a highly integrated formulation. Embedding the new scheme in a full, non-linear image registration algorithm enables fast calculations on high resolutions with dramatically reduced memory consumption. The new approach provides linear scalability compared with a traditional sparse-matrix-based scheme. The algorithm is evaluated on a challenging problem from radiotherapy, where pelvis cone-beam CT and planning CT images are registered. Speedups up to a factor of 149.3 for a single Hessian-vector multiplication and of 20.3 for a complete non-linear registration are achieved.

[1]  D K Smith,et al.  Numerical Optimization , 2001, J. Oper. Res. Soc..

[2]  H Handels,et al.  Lung registration using automatically detected landmarks. , 2014, Methods of information in medicine.

[3]  Jan Rühaak,et al.  Highly efficient image registration for embedded systems using a distributed multicore DSP architecture , 2014, Journal of Real-Time Image Processing.

[4]  Torsten Rohlfing,et al.  Nonrigid image registration in shared-memory multiprocessor environments with application to brains, breasts, and bees , 2003, IEEE Transactions on Information Technology in Biomedicine.

[5]  Jan Modersitzki,et al.  Curvature Based Image Registration , 2004, Journal of Mathematical Imaging and Vision.

[6]  Jürgen Teich,et al.  Frameworks for Multi-core Architectures: A Comprehensive Evaluation Using 2D/3D Image Registration , 2011, ARCS.

[7]  Jan Rühaak,et al.  A fully parallel algorithm for multimodal image registration using normalized gradient fields , 2013, 2013 IEEE 10th International Symposium on Biomedical Imaging.

[8]  Ludvig Paul Muren,et al.  Propagation of target and organ at risk contours in radiotherapy of prostate cancer using deformable image registration , 2010, Acta oncologica.

[9]  Steve B Jiang,et al.  Implementation and evaluation of various demons deformable image registration algorithms on a GPU. , 2010, Physics in medicine and biology.

[10]  L. König,et al.  A Non-Linear Image Registration Scheme for Real-Time Liver Ultrasound Tracking using Normalized Gradient Fields , 2014 .

[11]  Jan Rühaak,et al.  Highly accurate fast lung CT registration , 2013, Medical Imaging.

[12]  E. Haber,et al.  Intensity Gradient Based Registration and Fusion of Multi-modal Images , 2007, Methods of Information in Medicine.

[13]  Jan Modersitzky,et al.  FAIR - Flexible Algorithms for Image Registration , 2009, Fundamentals of algorithms.

[14]  Jürgen Teich,et al.  Frameworks for GPU Accelerators: A comprehensive evaluation using 2D/3D image registration , 2011, 2011 IEEE 9th Symposium on Application Specific Processors (SASP).

[15]  Hualiang Zhong,et al.  Assessment of dose reconstruction errors in image-guided radiation therapy , 2008, Physics in medicine and biology.

[16]  Lars Hömke,et al.  A multigrid method for anisotropic PDEs in elastic image registration , 2006, Numer. Linear Algebra Appl..

[17]  John D. Owens,et al.  Fast Deformable Registration on the GPU: A CUDA Implementation of Demons , 2008, 2008 International Conference on Computational Sciences and Its Applications.

[18]  Rodney A. Kennedy,et al.  A Survey of Medical Image Registration on Multicore and the GPU , 2010, IEEE Signal Processing Magazine.

[19]  Nassir Navab,et al.  A Novel Projection Based Approach for Medical Image Registration , 2006, WBIR.

[20]  Heinz-Otto Peitgen,et al.  GPU Accelerated Image Registration in Two and Three Dimensions , 2006, Bildverarbeitung für die Medizin.

[21]  Hao Wu,et al.  Fast wavelet‐based multiresolution image registration on a multiprocessing digital signal processor , 1998 .

[22]  Lars Konig,et al.  A fast and accurate parallel algorithm for non-linear image registration using Normalized Gradient fields , 2014, 2014 IEEE 11th International Symposium on Biomedical Imaging (ISBI).