Purchase this article with an account.
Jimmy Jiang Liu, Yanwu Xu, Beng Hai Lee, Min Thet Htoo, Damon Wong, Baskaran Mani, Tin Aung, Tien Wong; Architecture for Angle Closure Glaucoma: Novel methods for Imaging, Risk Assessment and Screening (AGAR). Invest. Ophthalmol. Vis. Sci. 2013;54(15):3575. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To introduce AGAR, a system to automatically identify glaucoma type with OCT (optical coherence tomography) images.
We developed an automatic system to identify glaucoma type with OCT (optical coherence tomography) images. Glaucoma is classified according to the configuration of the angle (the part of the eye between the cornea and iris mainly responsible for drainage of aqueous humor) into open angle (OA) and angle-closure (AC) glaucoma. The AGAR system is based on image processing and machine learning (Fig. 1), which differs from previous works that rely on anterior chamber angle (ACA) assessment metrics used clinically, such as AOD (angle-opening distance) [Leung, et al, Arch. Ophthalmol. 2006], TIA (trabecular-iris angle) [Leung, et al, Invest. Ophthalmol. Vis. Sci. 2008], TISA (trabecular-iris space area) [Tian, et al, TBME 2011] and SLBA (Schwalbe’s line bounded area) [Tian, et al, EMBC 2010]. The system was tested on 2048 high definition OCT images from the Singapore National Eye Centre, which were manually annotated by trained graders from the Singapore Eye Research Institute.
The 2048 images are from 8 circular scan videos of 8 patient eyes with glaucoma, 4 of them with PACG and other 4 with POAG. Each video contains 128 frames, and each frame is split into 2 images since it contains two angles. For the evaluation, we follow the widely used leave-one-out (LOO) method, i.e., for each testing round, 512 images from one PACG and one POAG patients are used for testing while others are used for training, thus 16 rounds are performed to test all cases. The proposed automated system achieves an AUC value of 0.921±0.036 and a combined accuracy of 84.0±5.7.
Using an innovative image processing and machine learning approach, AGAR system is able to achieve a good performance in angle closure assessment. The testing results on an extensive dataset are promising and demonstrate that the AGAR system has the potential to be expanded into an automated screening tool for primary angle closure glaucoma (PACG) detection.
This PDF is available to Subscribers Only