Porting a 3D image registration application to multi-core environment

摘要

Meanwhile distributed and parallel computing has been used for medical research applications, the most recent multi-core computer architectures bring along the opportunity to adapt these paradigms to the applied medical solutions as well, thus significantly enhancing their usability.rnWe have ported an already implemented 3D image registration algorithm to the Cell Broadband Engine platform. The algorithm is able to adjust and align digitized tissue slices cut from human body into one coordinate system. Our re-designed algorithm to this heterogeneous multi-core processor shortened the processing time by 70 %.rnUnlike the pictures available after an MRI or PET scan, these tissue slices are situated on the picture at different positions with different angles, significantly strained at random parts and might be disordered during the digital acquisition.rnThis paper is a report on how the Coarse Mutual Adjustment algorithm of the MIRAX 3D module of MIRAX Viewer [1], digital microscope software was ported to the Cell Broadband Engine. It introduces the application of the software, the previously developed algorithm, describes the porting phases, and gives performance measurement results and conclusions.