Comparison of smartphone-based fundus imaging and ultra widefield fundus imaging for retinopathy of prematurity screening at Aravind Eye Hospital

Abstract

Purpose : Telemedicine for remote grading of images has been effective in retinopathy of prematurity (ROP) screening in resource-limited settings. Ultra widefield fundus imaging systems can be cost-prohibitive for widespread use. Smartphone-based digital fundus imaging has been described as a low-cost alternative, however the relative effectiveness of this approach has not been fully evaluated. The purpose of this study is to compare ROP diagnosis between the Retcam (Natus Medical Incorporated, Pleasanton CA, USA) and a smart-phone based approach using the MII (Make in India) Retcam adapter.

Methods : 100 paired exams from 95 infants imaged on both the Retcam and MII were acquired during routine image-based ROP screening via telemedicine between August-September 2021 at Aravind Eye Hospital. Images were uploaded to a web-based data management system (iTeleGENx, Chicago, IL, USA) which enabled randomization and matching. A single grader was presented with demographics including birthweight and gestational age, and determined the ROP diagnosis according to the International Classification of Retinopathy of Prematurity. Exams were anonymized and presented in two time periods in random order to prevent recall bias. Sensitivity and specificity calculations were made for TR-ROP, Type 2 or worse, and pre-plus or worse in at least one eye.

Results : Using the Retcam as the gold standard, the sensitivity for type 2 or worse in either eye was 78.6% and specificity was 97.6%. For TR-ROP, the sensitivity was 100% and specificity was 98.9%. The sensitivity for Type 2 or worse on the Retcam and pre-plus or worse on the MII was 85.7% specificity was 90.6%. Table 1 summarizes the diagnosis of ROP on both devices with regards to zone, stage, presence or absence of plus disease, and category of disease.

Conclusions : The MII Retcam demonstrated high sensitivity and specificity for TR-ROP, though did not capture all eyes with type 2 ROP. This may be due to limitations in field of view, image quality, learning curve, or other factors. Our analysis suggests that if pre-plus can be clinically recognized using the MII Retcam, it could be used as a surrogate for the presence of Type 2 or worse disease. Referral criteria using lower field of view cameras should be evaluated, including potential applications of artificial intelligence to these cameras to evaluate vascular severity.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.