Research work · 2014

Accelerating adaptive ultrasound imaging algorithms by means of general-purpose computing on graphics processing units

Jon Petter Åsen

NORA - Norwegian Open Research Archives

Abstract

A rapid development in computer game technology and accompanying programming\nlanguages have recently provided researchers with small personal supercomputers,\ncomprised in a single graphics processing unit (GPU). This immense rise in\ncomputational capabilities and improved programmability are currently changing\nhow ultrasound imaging systems are designed. When researchers are exploring new\nalgorithms for ultrasound imaging, it is therefore important to keep the architecture\nof parallel accelerators like the GPU in mind. If a new complex algorithm is supposed\nto run in real time, it needs to fit the programmable and parallel pipeline of modern\nultrasound scanners.\nThe aim of this study has been to investigate the possibility of utilizing GPUs\nfor advanced processing in an ultrasound imaging system. Among the investigated\nproblems are both adaptive beamforming, adaptive visualization of ultrasound\nvolumes, and ultrasound simulations. The presented problems have in common that\nthey require parallel programming in order to reach real-time processing.\nIn the first part of the thesis, the Capon adaptive beamformer is investigated\nand implemented on a GPU for the application of real-time sonar (Paper I) and\nmedical ultrasound imaging (Paper II). Real-time frame rates are achieved for both\nmodalities. Paper II also presents, for the first time, videos where the Capon\nbeamformer has been applied on loops of simulated and in vivo medical ultrasound\nimages. In Paper III, we show that Capon beamforming does not provide shiftinvariant\nimaging in a real-time imaging setting. A method is then proposed that\nimproves the shift-invariant property. Shift-invariant imaging is essential if the method\nis ever to be used in practice.\nIn paper Paper IV we propose an adaptive method for visualization of volumetric\ncardiac ultrasound images. The method is capable of removing noise that by\nconventional methods would have occluded cardiac tissue. This work also shows\nthat with modern GPUs it is possible to add advanced visualization methods to an\nultrasound imaging system and still have real-time performance.\nFinally, we investigate how GPUs can be utilized to accelerate ultrasound\nsimulations (PaperV). The result of this work was a simulation program where\nultrasound array geometries can be interactively drawn and where the resulting\npressure field is simulated and visualized in real time.

Read free full text