High Performance Medical Image Computing

Presented by

Presented by Daniel Blezek, Biomedical Engineering, Mayo Clinic; Hans Peter Pfister, Harvard University; Eric Borisch, Mayo Clinic

Abstract

This tutorial provides an introduction to methodologies of high performance computing (HPC) applied to medical image processing and analysis. First, motivation for HPC will be presented, driven from real-world medical image examples. Software strategies for writing HPC code on commodity hardware is the first topic. The next topics present a broad introduction to available hardware for high performance computing, beginning with graphics processing units (GPU) and the Cell Broadband Engine (Cell) and concluding with cluster computing. A question and answer session will conclude the tutorial. On completion of this tutorial, participants will: 1) be able to make informed decisions as to the applicability of HPC to imaging problems, 2) have working knowledge of hardware and software HPC techniques.

The target audience for this tutorial is medical imaging researchers interested in understanding and evaluating HPC for imaging computing, including graduate students, post-doctoral fellows, and faculty. Participants are not required to have a computer science background, but are expected to have knowledge of common image processing algorithms and be familiar with simple programming, i.e. Matlab. Familiarity with C, C++ and/or Java helpful but not required.

With the explosion of multislice CT scanners, high-definition MRI and video streams from ultrasound devices, medical image computing demands new approaches capitalizing on high-performance computing capabilities available in today’s commodity components. The purpose of this tutorial is to present a broad overview and specific examples of high-performance computing applied to medical imaging. Following a brief introduction, four distinct areas of high-performance computing will be presented.

Topics will include: