Purchase this article with an account.
Xiang Wei, Acner Camino, Shaohua Pi, William Cepurna, David Huang, John C Morrison, Yali Jia; Real-time cross-sectional and en face OCT Angiography in Graphics Processing Unit. Invest. Ophthalmol. Vis. Sci. 2018;59(9):3925.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Defocusing, vignetting and bulk motion interfere with the image quality of optical coherence tomography angiography (OCTA) more dramatically than structural OCT. Assessment of the best focusing, alignment conditions and the stability of imaging subjects in the current OCTA systems is done without knowledge of the OCTA quality before acquisition, resulting in low yield rate for further quantification. We aim to develop a platform for real-time Split-Spectrum Amplitude-Decorrelation Angiography (SSADA) with real-time display of cross-sectional and en face images to assist operators during data acquisition and ensure scan quality.
We developed a Graphic Processing Unit (GPU)-based real-time signal and image processing program. Retinal OCTA scans were acquired from Brown Norway rats by a prototype visible-light spectral domain OCT system with 560-nm center wavelength and 50-kHz A-line rate. GPU was used to compute OCT reflectance and flow signal after each set of 3 B-frames were acquired at the same location and within the duration used for B-frames acquisition in the next position. The OCT reflectance amplitude and OCTA were computed by applying the multi-thread GPU-accelerated SSADA algorithm. Simultaneously, the mean projection of OCT reflectance and maximum projection of OCTA (en face images) were generated and displayed just after the scanning at each position. The software was programmed using C++ on a computer (Intel core i7 CPU, 32GB RAM, NVIDIA GTX1080ti GPU).
The processing speed of GPU-assisted OCTA was improved more than 100 times, compared to conventional OCTA processing using Matlab 2017a. The time used to generate single OCTA B-frames (512 × 2048 voxels) is about 15 ms (data transfer and calculation), which is just half of the time (30 ms) used for B-frame acquisition at each position. This technology enables real-time display of cross-sectional and en face OCT and OCTA during scan acquisition, for the first time. Cross-sectional flow signal and en face micro vasculatures can be easily appreciated on the OCTA display interface.
A real-time OCTA program was successfully developed and applied to retinal angiography in rodents. This provides a valuable tool for evaluating OCTA scans in real-time and providing feedback for adjusting OCT scan module position in order to improve OCTA scan yield rate by checking subject motion condition and avoiding vignetting and defocusing.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only