Abstract
Purpose: :
To evaluate the effect of digitally converted red–free monochromatic retinal nerve fiber layer (RNFL) images, by the non–mydriatic digital fundus camera and the image software, to detect RNFL defects.
Methods: :
Ninety–five eyes taken with both non–mydriatic digital fundus photography and conventional RNFL photography during diagnostic work–up for glaucoma were enrolled. The digitally converted red–free monochromatic RNFL images were acquired from non–mydriatic digital fundus photographs by removing red color and converting into black and white images using Photoshop 7.0 software (Adobe, San Jose). Inter– and intraobserver agreements and the presence of localized or diffuse RNFL defects were evaluated in the non–mydriatic digital fundus images and digitally converted red–free images. Regarding conventional RNFL photograph as a standard, the sensitivity, specificity, and positive and negative predictive values to detect RNFL defects were calculated in both the original and converted images.
Results: :
Interobserver agreements (Cohen’s kappa values) for the localized RNFL defects of the conventional RNFL images, non–mydriatic digital fundus images, and digitally converted red–free images were 0.749, 0.634, and 0.793, respectively, however, all were less than 0.6 with regard to the diffuse RNFL defects. For the localized RNFL defects, the sensitivity, specificity, and the positive and negative predictive values of the digitally converted red–free images were 85.7%, 95.9%, 85.7% and 94.1%, respectively, showing superiority to those of the non–mydriatic digital fundus images.
Conclusions: :
The localized RNFL defects can be detected more effectively when non–mydriatic digital fundus images were digitally converted into the red–free images.
Keywords: nerve fiber layer • imaging/image analysis: clinical