Non-rigid Registration for Large Sets of Microscopic Images on Graphics Processors

Antonio Ruiz, Manuel Ujaldon, Lee Cooper, Kun Huang
2008 Journal of Signal Processing Systems  
Microscopic imaging is an important tool for characterizing tissue morphology and pathology. Large sets of microscopic images are usually required for 3D reconstruction and visualization of tissue structure. Registration is essential for the 3D reconstruction from the stack of images. However, the large size of image datasets proves to be a challenge for automatic registration. In this paper, we present a novel method for completing this task more efficiently on graphics processors (GPUs) and
more » ... mbine its features with parallel programming to achieve speed-up factors of up to 4.11x on a single GPU and 6.68x on a pair of GPUs using CUDA and pthreads versus a typical C++ CPU implementation. Execution numbers are shown for a benchmark composed of large-scale images coming from two different sources: Genetic studies (16K × 16K pixels) and breast cancer tumors (23K × 62K pixels). It takes more than 12 hours for the genetic case in C++ to register a typical sample composed of 500 consecutive slides, which was reduced to less than 2 hours using two GPUs, in addition to a very promising scalability for extending those gains easily on a large number of GPUs.
doi:10.1007/s11265-008-0208-4 pmid:25328635 pmcid:PMC4198069 fatcat:kcwzubt6trdeleh4vq44jf36hy