Archives

Detecting Outer Edges in Retinal OCT Images of Diseased Eyes Using Graph Cut Method with Weighted Edges


Praveen Mittal and Charul Bhatnagar
Abstract

Segmentation of retinal layers in an OCT image is the first step towards automatic analysis of diseased retina. Various methods have been proposed to segment the seven retinal layers of normal eye. Methods have also been proposed to segment layers of retina inflicted with a particular disease. We propose a generalized method that is able to segment the outer layers of normal eye as well as that of eye having diseases like Drusen, Choroidal Neovascularization (CNV) or Diabetic Macular Edema. To segment the layers, Graph Cut Method is used. Each node of the graph is pixel in the image and weights are assigned for edges connecting these nodes. Finally shortest path from one node to another leads to layers of retina. We have tested our method on kermany dataset. The results were validated by two qualified ophthalmologists. We have got a root mean square error as 0.42 and dice coefficient as 0.03 for a set of 60 images including all categories of disease. The proposed method is able to permit for early diagnosis of macular diseases by the study of vital layer information for an efficient eye inspection.

Volume 12 | 03-Special Issue

Pages: 943-950

DOI: 10.5373/JARDCS/V12SP3/20201338