Accelerating Noninvasive Transmural Electrophysiological Imaging with CUDA

Martin Andrew Corraine
Department of Computer Engineering, Kate Gleason College of Engineering, Rochester Institute of Technology, Rochester, NY
Rochester Institute of Technology, 2012


   title={Accelerating Noninvasive Transmural Electrophysiological Imaging with CUDA},

   author={Corraine, M.A.},


   school={Rochester Institute of Technology}


Download Download (PDF)   View View   Source Source   



The human heart is a vital muscle of the body. Abnormalities in the heart can disrupt its normal operation. One such abnormality that affects the middle layer of the heart wall (myocardium) is called myocardial scars. Just like any tissue in the body, damage to healthy tissue will trigger scar tissue to form. Normally this scar tissue is benign. However, myocardial scars can disrupt the heart’s normal operation by changing the electrical properties of the myocardium. It is the most common cause of ventricular arrhythmia and sudden cardiac death. Leading edge research has developed a technique called Noninvasive Transmural Electrophysiological Imaging (NTEPI) to help diagnose myocardial scars. However, NTEPI is hindered by its high computational requirements. Due to the parallel nature of NTEPI, Graphics Processing Units (GPUs) equipped with the Compute Unified Device Architecture (CUDA) by Nvidia can be leveraged to accelerate NTEPI. GPUs were chosen over other alternatives because they are ubiquitous in hospitals and medical offices where NTEPI will be used. This project accelerated NTEPI with CUDA. First, NTEPI was profiled to determine where most of the time was spent. This information was used to determine what functions were chosen for CUDA acceleration. The accelerated NTEPI algorithm was tested for accurateness by comparing the outputs of the baseline CPU version to the CUDA version. Lastly, the CUDA accelerated NTEPI algorithm was profiled on three GPUs with different costs and features. The profiling was used to determine if any bottlenecks existed in the accelerated NTEPI algorithm. Lastly, CUDA specifications were identified from this profiling data to achieve the highest performance in NTEPI with and without cost as a factor.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: