Purchase this article with an account.
M.–L. Belair, M.C. Boucher, J. Martin; Conception and evaluation of a module for training readers in screening for diabetic retinopathy using non–mydriatic digital photography . Invest. Ophthalmol. Vis. Sci. 2004;45(13):5248.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Background: Non–mydriatic camera is a promising technology for screening of diabetic retinopathy. Mass screening generates a large amount of images that need to be analyzed efficiently, Training readers would facilitate the establishment of a screening program. Purpose: Part A: Development of educational software for preparation of a first level "Trained Reader" to detect abnormalities related to diabetic retinopathy. Part B: Validation of educational software by mesuring agreement between "Trained Readers" and ophthalmologists. Methods: Part A: The authors developed software illustrating the normal retina and the abnormalities related to diabetic retinopathy. Part B: Five individuals were asked to independently follow the training module. After completion of the training module, the individuals analyzed a group of images from the non–mydriatic digital camera. These images represented 100 eyes with different degrees of pathology. The "Trained Readers" had to decide for each group of two images (one centered on the optic nerve and one centered on the macula) between: presence of abnormality, absence of abnormality or insufficient quality of the images. The images had previously been analyzed by two ophthalmologists. Agreement between the readers and the ophthalmologists was then determined. Results: Mean percent agreement between readers and ophthalmologists was of 86.2% (95%CI 80.5% to 91.9%). Kappa coefficient, which corrects for chance agreement, was between 0.610 and 0.800 with a mean of 0.709. Based on conventional interpretation of kappa coefficient, value for the five readers were all in the range of good agreement. Conclusions: Authors believe that the training module could eventually be used to formally develop "Trained Readers" of digitized images and contribute to the establishment of mass screening programs for diabetic retinopathy.
This PDF is available to Subscribers Only