A Parallel Implementation of 2-D/3-D Image Registration for Computer-Assisted Surgery

Ino, Fumihiko; Kawasaki, Yuki; Tashiro, Taku; Nakajima, Yoshikazu; Sato, Yoshinobu; Tamura, Shinichi; Hagihara, Kenichi

doi:10.1109/icpads.2005.40

Cited by 4 publications

(7 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Threads have local memory that can be readily implemented on distributed memory platforms such as clusters. Using MPI, we distributed the gradient computation equally across nodes in a small cluster similar to [3]. This distribution is possible by virtue of the fact that each finite difference calculation for each control point is independent and requires only the neighboring voxels and control points to calculate.…”

Section: Exploiting Multiple Forms Of Parallelismmentioning

confidence: 99%

“…Optimization level parallelism represents those parts of an algorithm that can run in parallel given the basic unit is an iteration of the image registration routine such as gradient decent invocations from different starting points [3] or with a distributed genetic algorithm [8]. Volume level parallelism is a generalization of optimization level parallelism where the computational units operate on entire volumes like computing a multidimensional gradient on whole image volumes [3].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Towards a Heterogeneous Medical Image Registration Acceleration Platform

Plishker

Dandekar

Bhattacharyya

et al. 2007

2007 IEEE Biomedical Circuits and Systems Conference

View full text Add to dashboard Cite

Abstract-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 100x 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 subminute medical image registration capabilities.

show abstract

Section: Exploiting Multiple Forms Of Parallelismmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Towards a Heterogeneous Medical Image Registration Acceleration Platform

Plishker

Dandekar

Bhattacharyya

et al. 2007

2007 IEEE Biomedical Circuits and Systems Conference

View full text Add to dashboard Cite

show abstract

“…Ino et al [9] use this idea (which they call "speculative parallelism") to promote faster convergence in their time-critical registration application. Since the best optimization parameters are difficult to identify a priori, multiple instances of the same algorithm are launched with different parameters.…”

Section: Optimization-level Parallelismmentioning

confidence: 99%

“…For example, an optimization iteration could be pipelined (applying one trial transform to the moving image while generating another candidate transform). Ino et al [9] discuss the potential of "task parallelism" in accelerating the gradient computation of a rigid registration algorithm. This is possible, since independent finite difference calculations are done using the entire volume.…”

Section: Volume-level Parallelismmentioning

confidence: 99%

See 1 more Smart Citation

Utilizing Hierarchical Multiprocessing for Medical Image Registration

Plishker

Dandekar

Bhattacharyya

et al. 2010

IEEE Signal Process. Mag.

View full text Add to dashboard Cite

dvances in medical imaging technologies have enabled medical diagnoses and procedures that were simply not possible a decade ago. The accelerating speed of acquisition and the increasing resolution of images have given doctors more information, which is taken less invasively about their patients. However, because of the multitude of imaging modalities [e.g., computed tomography (CT), positron emission tomography (PET), magnetic resonance imaging (MRI), and ultrasound (US)] and the sheer volume of data being acquired, utilizing this new data effectively has become problematic. One way to tap into the potential of this raw data is to merge these images into one integrated view through a procedure called image registration. COMPUTE-ENABLED MEDICINEFor the past decade, improving performance and accuracy has been a driving force of innovation in automated medical image registration. 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 processing units (GPUs) and general-purpose multicore systems, but there has yet to be a solution fast enough and effective enough to gain widespread clinical use.

show abstract