MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
EmbargoRelease after 22-Mar-2018
AbstractChronic Heart Failure (CHF) affects millions of Americans each year and it is the leading cause of hospitalization of the patients over the age of 65. The 3D cardiac simulation based on the bidomain model play an important role in understanding the CHF by simulating the interactions between the tissue cells. However, the computation complexity of the cardiac models has led cardiac researchers to less accurate models that are computationally tractable. In this thesis, we propose algorithmic and Graphics Processing Unit (GPU) specific optimizations to significantly reduce the amount and the complexity of computations needed for solving the bidmoain model. We propose a data reduction strategy and 2D cut based workload partitioning strategy to minimize the data transfer overhead and achieve strong scalability when executing the simulations on a multi-GPU environment. We propose an autonomic management framework based on the physics aware programming (PAP) paradigm for accelerating the cardiac simulations of monodomain model further beyond what can be achieved through traditional parallelization efforts. We apply machine learning techniques to detect the phase of the simulation during each time step of the 3D model of a human cardiac simulation. We dynamically change the resolution of the simulation based on the detected phase, and adjust the numerical solvers to optimize performance without sacrificing the accuracy of the simulation. We show that the scalable, algorithmically optimized and PAP-based implementation of bidomain model on 16 GPUs reduces the execution time of the 3D cardiac simulations by a factor of 128 and 145124 compared to the state-of-the-art multi-GPU (on 16 GPUs) implementation and CPU-based serial implementation respectively based on tissue size of 256×256×256 with a simulation accuracy of 99.9%. This drastic reduction in simulation time will allow clinicians to accurately identify the CHF patients prone to ventricular arrhythmias, rapidly evaluate and develop therapy options, and perform interactive cardiac analysis without the risk of the invasive procedure.
Degree ProgramGraduate College
Electrical & Computer Engineering