High-speed FPGA-GPU processing for 3D-OCT imaging

Three dimensional (3D) imaging using optical coherence tomography (OCT) is equipped with a field-programmable gate array and graphics processing unit (FPGA-GPU) acquisition and processing architecture, thereby making it highly advantageous in the domain of parallel computing. To realize the full benefit of the data acquisition and processing capabilities, it is preferable to increase the number of high-speed processing modules capable of running the complicated image processing algorithms at comparable speeds. In this paper, we propose the design of a real-time image acquisition and pre-processing FPGA via LabVIEW (National Instruments (NI)) with GPU-based acceleration that is capable of sustaining the rate of data acquisition. When using the NI LabVIEW FPGA to develop high-speed processing, FPGA cores are modeled as reusable code modules and comprise subVIs, which are commonly implemented in the LabVIEW environment. The OCT pre-processing was implemented and performed via subVIs in LabVIEW FPGA. Additionally, we utilized GPU-based acceleration, which employs the use of a general purpose GPU (GP-GPU) together with a CPU, to accelerate the operation at hand. Finally, we implemented the LabVIEW FPGA core to perform data acquisition and image pre-processing in the frame grabber. Results showed that, by applying GPU acceleration to the tomographic inspection of biological samples, SD-OCT imaging in excess of 40 frames/s (FPS) for the NVIDIA M6000 GPU-accelerated SD-OCT with frame size 4096 (axial) × 512 (lateral) becomes feasible, and more than 512 × 512 × 500 volumes can be reconstructed with a speed increase of at least 7x that of a non-GPU.