Automated retinal layer segmentation of OCT images in normal and AMD eyes

Abstract

Optical coherence tomography (OCT) is a non-invasive, fast imaging technique that is widely used clinically for the diagnosis of ophthalmic diseases. It is very important to obtain quantitative retinal layer information, however, this approach is time consuming and challenging for ophthalmologists since it requires segmentation of the retinal layer in OCT images. An automated retinal layer segmentation method is proposed by employing N-sigmoid and complex diffusion filtering along with signal-noise ratio balance for pre-processing and fuzzy C-mean for clustering. Pre-processing increases the contrast between the retinal layers which eliminates the influence of speckle noise and blood vessels for later segmentation. The eigenvectors of each extremum were calculated and clustered by fuzzy C-means (FCM). The boundaries of each retinal layer were fitted using RANSAC and then retinal layer segmentation of the retina in the fundus OCT images was achieved. The proposed method can accurately obtain five retinal layers in OCT images affected by spackle noise, low image contrast and irregularly shaped structural features such as blood vessels.