Digital Ophthalmoscope Study

The Utility of a Smartphone-Based Retinal Imaging Device as a Screening Tool in an Outpatient Clinic Setting 

Principal Investigators: Siva Iyer MD & Mark Sherwood MD

Co-Investigators: Ajay Mittal, Victor Sanchez, Keer Zhang, Daniel Rodriguez

Purpose

In underserved areas, smartphone-based retinal screening has the potential to provide scalable, low-cost ophthalmic care. Telemedicine ophthalmoscopy usage in primary care clinics can provide benefits in identifying individuals that need further eye care. This study aims to determine capacity of this technology to record optic nerve and assess the potential of utilizing a smartphone ophthalmoscope, D-EYE, as a screening device for glaucoma. 

Methods

Over 150 enrolled participants were surveyed and had a 30-second video recording of their non-dilated eyes captured using D-EYE. A data set of 119 participants, 105 healthy and 14 glaucomatous, was graded independently by 5 resident ophthalmologists. Another dataset of 114 participants, 105 healthy and 9 glaucomatous, was graded by 4 other residents using the best captured frames from the 30 second videos to determine whether still images alone could provide a comparable result as the full video.

The grading criteria for both data sets included ability to visualize the optic nerve, subjective clarity measurements, cup-to-disc ratios, and likelihood of reporting patients for additional care.

Statistical analysis for both data sets include IRR and kappa coefficient using R 3.6.1 to measure agreement between graders. Kruskal-Wallis Rank Sum tests were conducted on average cup-to-disc ratios and likelihood of referral to compare the healthy and sick populations.

Results

In the 119 participant video-only dataset, the optic nerve could be visualized in 82.7% of cases. Mean reported cup-to-disc ratios were .299 for healthy participants (ICC=0.448) and .420 for those with +dx (ICC=0.341). Likelihood of reporting participants for additional care (graded on a scale of 1-5) was 1.67 in healthy  and 2.57 for +dx (ICC=0.509). Difference was statistically significant (p<.001). As cup-to-disc ratios increased for the sick population so did likelihood of referral. 

In the 114 participants still image-only dataset, the optic nerve was visualized in 76% of samples on average. Mean likelihood of referral by residents was poor at 1.8 for glaucoma and 2.13 for non-glaucoma (combined kappa= 0.15). This larger data set is being reanalyzed using the full video samples for efficiency comparison.

Conclusions

Current analysis indicates that the D-EYE may be useful for glaucoma screening. The data suggests a notable difference in the quality of data captured and referral recommendation between the full 30-second video and best still frames of the video. In addition, the data shows moderate agreement between raters for the likelihood of referral metric for the video data.