An image processing technique to detecting retina layers

In this paper, a computer approach is proposed for recognition of retina layers on optical coherence tomography (OCT) images. OCT uses the optical backscattering of light to scan the eye and describe a pixel representation of the anatomic layers within the retina. Our approach is based on co-occurrence matrix for feature extraction and a neural network for classifying, which four features of this matrix have been selected as a feature vector and multilayer perceptron (MLP) has been used for classifying retina layers. Achieved result of combined these two methods in the best state was 96.6% precision. This result shows that apply these methods on OCT images discriminate retina layers with efficient accuracy. Since, recognition of retina layers is important for automatic analysing of OCT images, therefore our proposed methods can be very useful.