Aims: To compare the results of fundus photography using a new non-mydriatic digital camera with the results of reference standard of Early Treatment Diabetic Retinopathy Study (ETDRS) retinal photographs, for the detection of diabetic retinopathy (DR).
Methods: Fundus colour photographs were taken with a Topcon non-mydriatic camera of 147 eyes of 74 diabetic patients, without pupillary dilation (five overlapping fields of 45 degrees; posterior pole, nasal, temporal, superior and inferior). Three retinal specialists classified the photographs in a masked fashion, as showing no DR or mild non-proliferative DR (NPDR) not requiring referral, moderate or more severe NPDR and/or macular oedema, or as non-gradable image requiring referral. ETDRS 35-mm colour slides served as reference images for DR detection.
Results: For moderately severe to severe DR, the sensitivities of detection reported by the three observers were 92, 100 and 92%, respectively, and the specificities, 87, 85, and 88%. For four levels of DR severity (none or mild NPDR, moderate NPDR, severe NPDR and proliferative DR), the percentages of exact agreement between the three observers on the retinopathy grades assigned to the non-mydriatic photographs and to the ETDRS reference slides were 94.6, 93 and 87.6%, respectively (kappa 0.60-0.80). Sixteen eyes of nine patients (11%) were judged ungradable by at least one observer. In a second series of 110 patients, evaluated in the setting of a screening procedure, fewer photographs were ungradable (< 6%).
Conclusion: These results suggest that fundus photographs taken by the Topcon TRC-NW6S non-mydriatic camera, without pupillary dilation, are suitable for DR screening.