Abstract
Purpose :
The goal was to produce a new fast, robust algorithm capable of segmenting retinal layers from OCT images. The algorithm is intended for use on the very large datasets and field of views acquired using a widefield Swept Source OCT prototype in clinical situations, for the study of eyes with significant pathologies. This is particularly important for applications in OCT angiographic imaging where visualization of blood flow within the appropriate anatomical regions is crucial for the appropriate clinical interpretation of images.
Methods :
A new fully automated algorithm was used to segment several retinal layers from both OCT intensity and OCT angiography datasets. In particular, we studied images obtained from a prototype 100-kHz SS-OCT instrument (Carl Zeiss Meditec, Dublin, CA) with a central wavelength of 1,050 nm. This instrument is capable of acquiring both intensity and angiography scans over retinal areas up to 12x12mm (512x512 to 1000x1000 cubes for intensity scans, 450x450 cubes for OCT angiography).
Results :
The segmentation algorithm generated slabs used for OCT angiography en face visualization. We obtained Inner Retinal (ILM to OPL), Outer Retinal (OPL to RPE, OPL to BM, RPE to BM), Choriocapillaris, and Choroidal slabs. Expert comparison with OCT angiography images obtained with a supervised semi-automated segmentation showed equivalent results. Total processing time was below 30s in a Matlab environment on a Dell Precision laptop.
Conclusions :
The algorithm produced a robust, fast segmentation on datasets from patients with a wide variety of retinal pathologies. It is shown to be useful for generating widefield OCT angiography slabs in a typical clinical setting.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.