# Performance of a New Tabletop Non-mydriatic Fundus Camera for Single-Field Diabetic Retinopathy Detection: A Pilot Study

## TL;DR

A new tabletop fundus camera was tested for detecting diabetic retinopathy and showed performance similar to standard devices.

## Contribution

The study evaluates a novel non-mydriatic tabletop fundus camera for diabetic retinopathy detection with a single-field imaging approach.

## Key findings

- The tabletop camera detected DR in 23.86% of patients, similar to the standard camera's 23%.
- The camera showed high inter-modality agreement (kappa 0.927) and strong sensitivity and specificity for DR detection.
- The tabletop camera had a slightly higher percentage of usable images compared to the standard camera.

## Abstract

Introduction: Diabetic retinopathy (DR) detection is made easy with the use of a fundus camera. The evidence of the use of a fundus camera for DR detection in non-mydriatic conditions with limited technical challenges is scarce. This is a pilot study that evaluates the performance of the Oivi fundus camera (Oivi AS, Oslo, Norway), a novel non-mydriatic tabletop fundus camera for DR detection using a single-field, macula-centered imaging approach. Its diagnostic accuracy was compared with that of a standard reference device, the Topcon NW400 fundus camera (Topcon Corporation, Tokyo, Japan).

Methods: A total of 243 subjects with diabetes mellitus (DM) were recruited. Non-mydriatic macula-centered images were captured using both cameras. Two ophthalmologists independently graded the deidentified images for image quality and DR stage. Discrepancies between their assessments were adjudicated by consensus after review by a senior ophthalmologist. The senior ophthalmologist's grading of images from the standard camera images served as the ground truth for comparative analysis. Inter-modality agreement was evaluated using linear weighted kappa (κ) correlation.

Results: DR was detected in 23% of patients using the standard reference camera (12% of eyes) and in 23.86% using the tabletop camera (12.6%). Identification of moderate non-proliferative diabetic retinopathy (NPDR) (7.2%), severe NPDR (0.4%), and proliferative diabetic retinopathy (PDR) (1.23%) was similar between the two cameras, although not always in the same eyes. The inter-modality agreement (k) for DR was 0.927 (95% CI: 0.88-0.97) (almost perfect). The tabletop camera showed a sensitivity of 92.98% (95% CI, 83-98.05%) and a specificity of 99.47% (95% CI, 98.10-99.94%) for DR. The percentage of usable images was 92.3% with the standard reference camera and 95.2% with the tabletop camera.

Conclusions: This study provides preliminary evidence that the novel tabletop Oivi fundus camera may offer comparable performance to standard non-mydriatic devices for DR detection in a single-field strategy. Its portability and usability under mesopic conditions suggest potential value for point-of-care screening. Further large-scale studies are warranted to validate these findings and explore their role in screening programs.