Abstract
Purpose :
Telemedicine may play a key role in improving screening for diabetic retinopathy (DR), but its scope is limited by the traditional use of fixed cameras. The present study aims to compare clinical assessment of referral-warranted DR by standard, nonmydriatic, fixed fundus camera photography with data gathered using a smartphone-based store-and-forward tele-ophthalmology platform.
Methods :
This was a clinic-based, cross-sectional, comparative instrument study. 120 patients at Santa Clara Valley Medical Center, a healthcare safety-net county hospital, underwent screening for diabetic retinopathy employing standard nonmydriatic, fixed camera fundus photography (Nidek NM-1000) and dilated fundus photography employing a smartphone camera with lens adapter (DigiSight Scope). Near spectacle-corrected visual acuity and nonmydriatic anterior segment photos to determine the presence of iris rubeosis were also captured using the smartphone platform (DigiSight SightBook). Masked graders performed standardized grading of nonmydriatic fixed camera-acquired and mydriatic smartphone-acquired photos. Quantitative comparison of fixed camera and smartphone-acquired data was performed using descriptive, kappa, Bland-Altman, and receiver operating characteristic analyses.
Results :
We report on a preliminary analysis of the first 65 patients enrolled in the study. Smartphone visual acuity was successfully measured in all eyes. Mean visual acuity of right and left eyes measured 70±13 and 65±16 ETDRS letters, respectively. All anterior segment photos were of sufficient quality to grade, and two eyes (3%) were found to have iris neovascularization. Fixed camera and smartphone-acquired fundus photos were of sufficient quality to grade in 59 (91%) and 61 (94%) eyes, respectively. Smartphone-acquired fundus photos demonstrated 98% sensitivity and 40% specificity to detect moderate nonproliferative and worse diabetic retinopathy, with good agreement between fixed camera and smartphone-acquired photo grades (kappa=0.87±0.1, p<0.001; AUROC=0.87, 95% confidence interval, 0.74-0.99).
Conclusions :
We report on a smartphone-based telemedicine system that demonstrated sensitivity in a healthcare safety-net setting to detect referral-warranted DR as compared with standard in-clinic fundus photography; thus, it may be useful in expanding DR screening to aid in preventing blindness through the timely detection and treatment of DR.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.