Segmentation of the retinal layers in OCT images is the critical step in analyzing OCT volumetric data for diagnosis and monitoring of retinal disease progression. Real-time retinal layer segmentation has become increasingly desirable with the increasing OCT acquisition speed. In this work we explored methods to accelerate image processing method to segment retinal layers in OCT B-scan images including graph cut and deep learning. We demonstrated ~30-ms and ~3-ms segmentation of 7 retina layers per OCT B-scan with graph cut and deep learning respectively. The accelerated OCT B-scan segmentation was then integrated with our GPU OCT image acquisition software.
Sensorless Adaptive Optics (SAO) allows easy integration of adaptive optics in retina imaging systems, however the iterative nature of the SAO optimization process requires long time to perform aberration correction and the inevitable subject motion during the optimization could compromise the AO correction. Here we present a multi-modal SAO retina imaging system that includes Optical Coherence Tomography (OCT), OCT-Angiography, confocal Scanning Laser Ophthalmoscopy (cSLO), and fluorescence detection. To mitigate the motion artifact and increase the SAO performance, we developed volumetric image tracking to extract merit function of SAO only within the region of interests.