GPU-Accelerated Interpolation for 3D Image Registration

Authors: Naveen Himthani (University of Texas, Institute for Computational Engineering and Sciences), Andreas Mang (University of Houston), Amir Gholami (University of California, Berkeley), George Biros (University of Texas, Institute for Computational Engineering and Sciences)

Abstract: Image registration is a key technology in image computing with numerous applications in medical imaging. Our overarching goal is the design of a consistent and unbiased computational framework for the integration of medical imaging data with simulation and optimization to support clinical decision making for glioma brain tumor patients. A major issue in 3D image registration is the time to solution, which poses the demand for effective numerical schemes and the utilization of high performance computing resources.

In this poster, we extend present a GPU-accelerated implementation of the Lagrange interpolation kernel using hardware texture filtering feature of modern hardware. Ongoing work involves implementing a unified single-GPU code for 3D image registration along with other computational kernels such as FFT. I will present my work by briefly explaining image registration followed by the explanation of the interpolation algorithm and its features and then demonstrate the results obtained.

Best Poster Finalist (BP): no

Poster: pdf
Poster summary: PDF


Back to Poster Archive Listing