The autofluorescence (AF) derived from the retinal pigment epithelium has the potential to be used as a feature for early glaucoma diagnosis. The aim of this paper is to design and evaluate the method used for semi-automatic segmentation of the zones with increased autofluorescence. A randomized sample of 20 patients (age: 56 +/- 10 years) with open angle glaucoma was used. A new method for semi-automatic detection and segmentation of the zones with a higher level of autofluorescence in the junctional zone of parapapillary atrophy has been evaluated. Good agreement has been observed between manually and semi-automatically segmented zones for most of the images, but higher differences may occur for larger hyperfluorescent regions. The data were evaluated using the limits of agreement, with a 2sigma border. The described method allows fast and objective evaluation of AF images, preventing undesirable inter-expert variance. A large proportion of AF images could be evaluated successfully by the developed procedure, and the obtained results are comparable to the manual procedure in most cases.

Department of Biomedical Engineering Brno University of Technology Kolejni 4 612 00 Brno Czech Republic