Performance of a Retinal Imaging Camera With On-Device Intelligence for Primary Care: Retrospective Study

Area of Science:

• Ophthalmology
• Medical Imaging
• Primary Care Technology

Background:

• Access to early diabetic retinopathy (DR) screening remains a significant barrier to timely detection.
• Operational challenges like cost and workflow integration hinder the adoption of retinal camera systems in U.S. primary care.

Purpose of the Study:

• To develop and evaluate a novel retinal screening system designed for seamless integration into primary care workflows.
• To assess the performance and usability of the Verily Numetric Retinal Camera (VNRC) in a simulated primary care environment.

Main Methods:

• Development of a nonmydriatic, 45° field imaging retinal camera (VNRC) with on-device intelligent features and cloud-based image routing.
• Retrospective performance evaluation comparing VNRC against a reference camera and assessing its quality control algorithm's correlation with gradability.
• Usability study involving trained and untrained users and operators in a simulated primary care setting, followed by user experience questionnaires.

Main Results:

• The VNRC captured images with 98.5% sufficiency for clinical interpretation, comparable to a reference camera (97.1%).
• The VNRC's quality control algorithm showed a strong positive association (φ=0.58) with ophthalmologist-determined gradability.
• All users (100%) and most operators (97%) found the VNRC imaging process intuitive, comfortable, and a positive experience.

Conclusions:

• The VNRC system demonstrates strong performance and usability, supporting its implementation as an integrated retinal screening service in primary care.
• Further studies are recommended to validate real-world usability across diverse primary care settings.