Accurate automatic segmentation of retina layers with emphasis on first layer

TL;DR

This paper presents an improved automatic segmentation method for retinal layers in OCT images, emphasizing accurate detection of the first layer crucial for diagnosis, using enhanced algorithms to confine search areas and improve precision.

Contribution

The study introduces additional steps to traditional segmentation methods, specifically confining the search area for the first retinal layer to improve detection accuracy.

Findings

High accuracy in segmenting all retinal layers

Enhanced detection of the first layer near the fovea

Effective confinement of search area improves segmentation precision

Abstract

Quantification of intra-retinal boundaries in optical coherence tomography (OCT) is a crucial task for studying and diagnosing neurological and ocular diseases. Since manual segmentation of layers is usually a time consuming task and relay on user, a lot of attempts done to do it automatically and without interference of user. Although for extracting all layers usually same procedure is applied but finding the first layer is usually more difficult due to vanishing it in some region specially close to Fobia. To have a general software, beside using common methods like applying shortest path algorithm on global gradient of image, some extra steps are used here to confine search area for Dijstra algorithm especially for the second layer. Results demonstrates high accuracy in segmenting all present layers, especially the first one that is important for diagnosing issue.