GPU在眼科FD-OCT系统实时图像显示及数据处理中的应用

Applying Graphics Processing Unit in Real-time Signal Processing and Visualization of Ophthalmic Fourier-domain OCT System

将基于图像处理单元(GPU)的统一计算设备架构(CUDA)引入到眼科OCT系统成像的数据处理过程中,可实现并行数据处理,用CUDA对相关的运算及其算法进行优化,解决了目前影响眼科OCT系统实时成像的技术瓶颈。实验结果表明,将GPU作为通用并行计算处理器,采用GPU+CPU模式实现成像数据处理的速度,较传统的基于CPU平台的串行计算和成像模式执行同样数据处理的速度提高超过一个数量级,达到了临床2D实时成像的要求。

This investigation introduces GPU （Graphics Processing Unit）- based CUDA （Compute Unified Device Architecture） technology into signal processing of ophthalmic FD-OCT （Fourier-Domain Optical Coherence Tomography） imaging system, can realize parallel data processing, using CUDA to optimize relevant operations and algorithms, in order to solve the technical bottlenecks that currently affect ophthalmic real-time imaging in OCT system. Laboratory results showed that with GPU as a general parallel computing processor, the speed of imaging data processing using GPU＋CPU mode is more than dozens times faster than traditional CPU platform based serial computing and imaging mode when executing the same data processing, which reaches the clinical requirements for two dimensional real- time imaging.