Approaches to General Purpose GPU Acceleration of Digital Signal Processing in Optical Coherence Tomography Systems

This paper explores and evaluates two approaches designed to improve the performance of digital signal processing within optical coherency tomography systems. Such systems rely on high-speed cameras or fast photo detectors to capture data, that is then processed using established techniques (typically Fourier transforms) and results presented to the user. In certain applications of these systems, medical imaging in ophthalmology for instance, performance is an issue. CPU processing cannot keep pace with data capture, resulting in lost frames and longer latencies from scan to results. To address these issues, this research explores the use of commonly available graphics processors for the bulk of data processing in such systems. We have integrated this within an existing system, developed using National Instruments' LabVIEW software, and report on the results.