Fast Deformable Registration on the GPU: A CUDA Implementation of Demons

Abstract

In the medical imaging field, we need fast deformable registration methods especially in intra-operative settings characterized by their time-critical applications. Image registration studies which are based on Graphics Processing Units (GPUs) provide fast implementations. However, only a small number of these GPU-based studies concentrate on deformable registration. We implemented Demons, a widely used deformable image registration algorithm, on NVIDIA's Quadro FX 5600 GPU with the Compute Unified Device Architecture (CUDA) programming environment. Using our code, we registered 3D CT lung images of patients. Our results show that we achieved the fastest runtime among the available GPU-based Demons implementations. Additionally, regardless of the given dataset size, we provided a factor of 55 speedup over an optimized CPU-based implementation. Hence, this study addresses the need for on-line deformable registration methods in intra-operative settings by providing the fastest and most scalable Demons implementation available to date. In addition, it provides an implementation of a deformable registration algorithm on a GPU, an understudied type of registration in the general-purpose computation on graphics processors (GPGPU) community.

Evaluation of Free Form Deformation and Demons Registration with Discontinuities

“…During the time, several improvements regarding the precision and registration speed were developed. In Muyan-Ozcelik [4], GPU implementation of Demons algorithm was developed. Image registration is a high time consuming operation.…”

A GPU framework for parallel segmentation of volumetric images using discrete deformable models

“…We can find in the literature GPU-based implementations of the watershed [34] and region growing methods [32], along with Markov random fields [43] and graph cuts approaches [24,17]. Image registration also turned out to strongly benefit from parallel computing, giving birth to a variety of implementations based on mutual information [37], sum of squared differences [14], demons [31,20], viscous-fluid regularization [23] or regularized gradient flow [35].…”

Parallel reconstruction of neighbor-joining trees for large multiple sequence alignments using CUDA

“…6) is greater than MidPoint, it is mapped to coordinate (n+1-8, n+1-6) = (2, 4). Coordinate (7,2), it is not modified because the y coordinate (i.e. 2) is less than MidPoint.…”

“…The general-purpose computer with programmable GPU hardware has shown great promise for solving many computationally intensive problems and has opened up a range of possibilities for a variety of application domains ranging from scientific computing [3], computational geometry [4], database operations [5], image processing [6] [7], bioinformatics [8]and so forth. Earlier development on GPUs required that applications use explicit graphics application programming interfaces (APIs) to organize data into streams and invoke kernels, which were usually written using high-level programming languages, such as Cg, GLSL, etc.…”