Abstract
Purpose: :
Assessment of the inter-observer agreement in classification of Diabetic Retinopathy (DR) through digital retinography and RTA fundus images.
Methods: :
The fundus images of 30 patients of type II Diabetes Mellitus were included in the study. Each subject underwent an internistic evaluation and a complete ophthalmologic examination. Exclusion criteria were: refractive errors higher or equal 7 dioptres, concomitant ocular disorders, optical media opacities affecting image quality and visual acuity less than 20/80. Parallel observer-blind evaluations of the degree of DR, as defined on digital retinographies (TRC-50 IX Retinal Camera, Topcon Corp., Japan) and RTA (Retinal Thickness Analyzer, Talia Technology Ltd, Neve - Ilan, Israel) fundus images, were performed by three ophthalmologists (MM, MB, SL). The only variable measured was the degree of DR, according to the modified Early Treatment Diabetic Research Study (ETDRS) classification: the retinopathy could be classified as non-proliferative (moderate, severe and very severe NPDR), proliferative (early and high risk PDR), non diabetic retinopathy (NDR) or not classifiable (NC). The images were acquired in pharmacological mydriasis (higher or equal 7 mm). Kappa statistic was used to assess the agreement between techniques. The Κ statistic was calculated interpreting as Κ<0 indicating no agreement; Κ = 0.0 to 0.19, poor; Κ = 0.20 to 0.39, fair; Κ = 0.40 to 0.59, moderate; and Κ = 0.60 to 0.79, substantial; and Κ = 0.80 to 1.0, almost perfect agreement.
Results: :
When the entire group was divided into two subgroups based on the DR stage (NPDR vs PDR), no difference was found between RTA and Topcon images among the three users and Kappa statistic showed a perfect agreement. However, when the entire group was divided on the basis of ETDRS classification, the agreement decreased to fair (kappa = 0.43).
Keywords: diabetic retinopathy • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical