Skip to Main Content
For the past decade, improving the performance and accuracy of medical image registration has been a driving force of innovation in medical imaging. Accurate image registration enhances diagnoses of patients, accounts for changes in morphology of structures over time, and even combines images from different modalities. The ultimate goal of medical image registration research is to create a robust, real time, elastic registration solution that may be used on many modalities. With such a computationally intensive and multifaceted problem, researchers have exploited parallelism at different levels to improve the performance of this application, but there has yet to be a solution fast enough and effective enough to gain widespread clinical use. To achieve real time elastic registration, an implementation must simultaneously exploit multiple types of parallelism in the application by targeting a heterogeneous platform whose computational components (e.g. multiprocessors, graphics processors, field programmable gate arrays) match these types of parallelism. Our initial experiments indicate that an 8 node heterogeneous cluster can realize over 100times speedup compared to a high performance uniprocessor system. By creating a platform based on modern hardware, we believe that a heterogeneous compute platform customized for image registration can provide robust, scalable, cost effective sub-minute medical image registration capabilities.