Towards systematic exploration of tradeoffs for medical image registration on heterogeneous platforms

Abstract

Abstract-For the past decade, improving the performance and accuracy of medical image registration has been a driving force of innovation in medical imaging. The ultimate goal of accurate, robust, real-time image registration will enhance diagnoses of patients and enable new image-guided intervention techniques. With such a computationally intensive and multifaceted problem, improvements have been found in high performance platforms such as graphics processors (GPUs) and general purpose clusters, but there has yet to be a solution fast enough and effective enough to gain widespread clinical use. In this study, we examine the differences in accuracy and speed of implementations of the same image registration algorithm on a general purpose uniprocessor, a GPU, and a cluster of GPUs. We utilize a novel domain specific framework that allows us to simultaneously exploit parallelism on a heterogeneous platform. Using a set of representative images, we examine implementations with speedups of up to two orders of magnitude and accuracy varying from sub-millimeter to 2.6 millimeters of average error.

Large Scale Image Registration Utilizing Data-Tunneling in the MapReduce Cluster

“…Trade-offs analysis is important [1] for evaluating and benchmarking the design models of Big image registration. However, existing works focus mainly on implementation performance with the cluster resource allocation [18,26] and hardware acceleration. To fill the gap, we focused on both the algorithmic complexity and design gain with the distributed MapReduce framework along with the resource allocation.…”

Utilizing Hierarchical Multiprocessing for Medical Image Registration

“…Based on these results, we discuss the possibilities and the challenges of combining acceleration approaches that utilize complementary types of parallelism. This article expands on preliminary work on this subject, which is presented in [1] and [2].…”

Algorithm/Architecture Co-Exploration of Visual Computing on Emergent Platforms: Overview and Future Prospects

“…The principle of vision technologies in medicine is the same as in other areas: the extraction of visual information from image data for medical diagnosis. For example, computer tomographic image reconstruction and medical image registration are among the very active topics of computer vision for medical applications [31]- [33]. Along with other tasks in computer vision, applications in medical imaging have been regarded as computation-intensive tasks because of the large size of the test-data sets used (in tens of gigabytes) and of the highly detailed low-level features that must be extracted from medical images acquired from different energy modalities.…”