Abstract
Purpose :
To present an automated geographic atrophy (GA) detection algorithm for quantifying GA area as seen on swept source optical coherence tomography (OCT) scans of patients with either uni-focal or multi-focal GA.
Methods :
The detection algorithm is based on integrated attenuation, in which the signal intensities of the RPE complex are normalized by the chorioscleral signal as part of the calculation [1]. A series of image processing steps, including preprocessing, auto-thresholding and region growing, are applied to the 2D integrated attenuation map generated from a 3D volume for a result of binary GA region map. A post-processing procedure follows to refine the GA region. To test the accuracy, the manually delineated GA regions in corresponding fundus auto fluorescence (FAF) images were compared to the automatic detection results of volumes acquired from macular and wide 3D scans from a one-micron wavelength swept-source device (DRI OCT Triton, Topcon Corp., Tokyo, Japan). To test the repeatability, multiple 3D scans of the same GA eye were detected and intra-class correlation coefficients and the mean of standard deviations were calculated.
Results :
The currently developed GA detection method demonstrated the ability to reliably delineate GA with good repeatability. The GA areas that were automatically detected in OCT were found to be correlated to the GA area that was manually delineated in FAF.
Conclusions :
The proposed automated GA detection method was able to produce reliable results and is promising towards practical clinical usage in GA area quantification and progression tracking.
Reference[1]: Yang, Q. and Reisman, C. United States patent US9,526,412.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.