Automated Segmentation System for Optical Coherence Tomography Images of the Anterior Chamber

Abstract

Purpose:: Among other clinical applications, Optical Coherence Tomography (OCT) is now used to see structures in the eye’s anterior segment that provide assessment of the risk, onset and progression of primary open angle glaucoma (POAG). However, such assessment, today, is qualitative and subjective. More objective assessment is afforded by automated segmentation and analysis. In this work algorithms have been developed to automatically segment the anatomical structures of interest in assessing POAG.

Methods:: Algorithms were developed for segmenting the corneal surfaces, the iris, the lens, the connection point where the cornea meets the iris, and the iris pigment. The geometry of these structures provide information about the risk of glaucoma. The algorithms were implemented in C++ and the segmentation takes 1.5 seconds on average for each image on a Pentium IV machine with 1 GB memory.

Results:: The algorithms were tested with a test database consisting of 381 images obtained from many patients. Each image was segmented and the results overlain on the original image for verification. These verifications and other tests show that it is realistic to make automated quantitative clinical measurements that help assess the risk of glaucoma onset and progression. Such measurements include the Anterior Chamber Angle (ACA), Anterior Recess Area (ARA), Anterior Chamber Depth (ACD), among others.

Conclusions:: The described segmentation system helps to automatically detect the different structures of clinical interest and enable measurements which are otherwise time-consuming and prone to inconsistency and errors if done manually. The described system can form a valuable additional diagnostic resource for the clinician and researcher by enabling accurate, reliable and quantitative measurements.

Keywords: imaging/image analysis: clinical • anterior chamber • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)