Automated segmentation and extraction of posterior eye segment using OCT scans

TL;DR

This paper introduces an automated, multi-phase method for segmenting and extracting the posterior eye segments from OCT scans, achieving high accuracy in both healthy and diseased eyes.

Contribution

The novel approach combines adaptive thresholding and structure tensor techniques for precise segmentation of eye compartments in OCT images.

Findings

Achieved mean IoU of 0.874

Obtained mean Dice coefficient of 0.930

Validated on 1000 OCT scans

Abstract

This paper proposes an automated method for the segmentation and extraction of the posterior segment of the human eye, including the vitreous, retina, choroid, and sclera compartments, using multi-vendor optical coherence tomography (OCT) scans. The proposed method works in two phases. First extracts the retinal pigment epithelium (RPE) layer by applying the adaptive thresholding technique to identify the retina-choroid junction. Then, it exploits the structure tensor guided approach to extract the inner limiting membrane (ILM) and the choroidal stroma (CS) layers, locating the vitreous-retina and choroid-sclera junctions in the candidate OCT scan. Furthermore, these three junction boundaries are utilized to conduct posterior eye compartmentalization effectively for both healthy and disease eye OCT scans. The proposed framework is evaluated over 1000 OCT scans, where it obtained the mean intersection over union (IoU) and mean Dice similarity coefficient (DSC) scores of 0.874 and 0.930, respectively.