Feature extraction method for neuro-sensory retinal layer segmentation using statistical estimation in optical coherence tomography

We have developed statistical estimation based feature extraction methods for layer segmentation of neuro-sensory retinal images obtained from optical coherence tomography. For clinical diagnosis purposes, a compact functional layer differentiation is targeted in this system so that an upgraded model for the statistical edge detector is considered. Initially, by iteratively searching the maximum edges in regular scopes of A-scans of the image, rough locations of interfaces are found. Then, assigning locational information in sequence to the detected edges, the interfacial locations are accurately detected. The proposed system has been successfully developed for identifying eight retinal layers and the accuracy is much comparable to the commercial equipment. With progressive improvement, we believe that this system will extensively provide the most practical application for various quantitative analyses in clinical diagnoses.