Fast 3D rigid medical image registration based on GPU
Wufan Chen · Jisuanji yingyong yanjiu · 2010
Automatic 3D image rigid registration technique,which maps multiple images into a common coordinate space,is an important clinical image analysis tool.The state of art rigid registration techniques like FLIRT take 300s to complete a rigid registration for a 2563 data set.This makes it hard to apply this technique in time critical clinical environment.This paper proposed a parallel accelerating registration algorithm based on the new CUDA.By using the GPU to compute the coordinate transformation,interpolation and similarity metric computation,great improvement of speed has been achieved without compromising the registration accuracy.The proposed technique is valuable for time critical clinical application.