Abstract
Purpose :
Glaucoma is known as the “silent thief of vision” due to its ability to cause debilitating visual deficits with few symptoms, a challenge exacerbated by a lack of objective diagnostic criteria. Optical coherence tomography angiography (OCTA) potentially allows for diagnosis before significant vision loss occurs by identifying glaucoma related changes in retinal and optic nerve head (ONH) microvasculature. Clinician OCTA analysis has been hampered by the complexity of the retinal blood supply; however, image processing programs such as the open-source software ImageJ now allow for objective, automated analysis of these complex images. The purpose of this study, therefore, was to assess whether ImageJ analysis of OCTA scans can match clinical data models at differentiating healthy and glaucomatous eyes.
Methods :
This IRB-approved, age-matched, retrospective study analyzed data from 85 healthy and 81 primary open-angle glaucoma (POAG) eyes diagnosed at an academic eye center. Clinical data often used in glaucoma diagnosis such as visual acuity, intraocular pressure, cup/disc ratio, central corneal thickness, and retinal nerve fiber layer thickness were collected for all eyes. OCTA scans of the macula and ONH were then analyzed by the image processing software ImageJ for 50 healthy and 50 POAG eyes. This analysis obtained numerical parameters describing the retinal vasculature, including vessel density (VD), vessel length density (VLD), and fractal dimension (FD). Logistic regression was performed on the ImageJ data and clinical data separately, creating two different diagnostic models that were then tested on a separate set of eyes not used for training (35 healthy, 31 POAG). Receiver operating curves were constructed to assess the models’ utility in differentiating healthy and POAG eyes.
Results :
VD, VLD, and FD were all significantly reduced in POAG compared to healthy eyes (p < 0.0001), particularly around the ONH. This led to no significant difference in the ability of the ImageJ model (AUC = 0.917, CI: 0.847-0.986) to differentiate between healthy and POAG eyes compared to the clinical data model (AUC = 0.910, CI: 0.841-0.978), with the ImageJ-based model performing slightly better.
Conclusions :
Our pilot study indicates that studying relatively objective OCTA vascular parameters using ImageJ may be equally effective to evaluating clinical features in diagnosing glaucoma.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.