Abstract
Purpose :
To determine if an Artificial Intelligent (AI)/ machine learning (ML) platform can learn and recognize optic nerve and nerve fiber layer changes in primary open angle glaucoma patients.
Methods :
This is an ongoing phase 2 study of the Retina Metrics – Synthetic Analyst AI platform. 500 patients with primary open angle glaucoma changes who had received ≥3 months of topical therapy in the preceding 18 months were analyzed. Topical medications had to have been previously prescribed and patients had to have a minimum period of follow up for at least 12 months. Therapies could be adjusted based on qualitative optical coherence tomography changes and or physician discretion. Changes in VA (Early Treatment Diabetic Retinopathy Study [ETDRS] letters), Nerve Fiber Layer thickness (NFL), microperimetry, ERG, VEP and Visual fields could also be assessed and clinically correlated. The last-observation-carried-forward method was used to impute missing data. Safety monitoring includes evaluation of ocular and systemic adverse events.
Results :
500 patients, VA and intraocular pressures were monitored and target IOP’s were established. The mean values and delta changes in both were graphed and trends lines were established. The NFL OCT, Imagery and quasi-imagery data were evaluated and assessed for the potential correlations. The ability for the AI platform to measure and learn appeared to improve as the data set enlarged. Means measurements improved by a mean (± standard error of the mean [SEM]) of 3.4 ± 0.7 and 5.8 ± 0.7 at initial testing. Current up to date data will be presented, as the AI/ ML platform appears to have the ability to improve its accuracy while learning from enlarging data sets.
Conclusions :
Primary Open Angle Glaucoma is a complex multifactorial disease process. The Retina Metrics – Synthetic Analyst AI platform, appears to have the ability to follow monitor changes in the optic nerve and nerve fiber layer. Monthly evaluations appeared to allow for improved IOP target control leading to a slower rate of potential VA and visual field loss. With larger data sets, longer periods of historic comparisons and further clinical correlation, this platform might allow for better predictive assumptions, aiding further therapy.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.