Abstract
Abstract: :
Purpose:To separate early glaucoma cases from normal subjects using an improved model of optic nerve head (ONH) topography, and without prior manual outlining of the optic disc or construction of a reference plane. Methods: Topography was modeled by a second–order surface to describe the peripapillary RNFL combined with a Gaussian Cumulative Distribution Function (GCDF) to describe the disc. Parameters derived from the best fit of the model to the image were used to classify eyes as normal or glaucomatous. The method was tested with 100 eyes of 100 normal subjects and 100 eyes of 100 subjects with early glaucomatous visual field damage (average MD = 3.24 dB). Results: The highest specificity (90%) and sensitivity (75%) values were obtained by mapping global model parameters to an individual glaucoma probability score via a Bayesian classifier. Similar results were obtained with a sectorally resolved model, which had a specificity of 85% and a sensitivity of 74% in the temporal inferior sector. Conclusions: The HRT, combined with an automated classification method provides good separation between control subjects and patients with early glaucoma.
Keywords: optic disc • image processing • imaging/image analysis: non-clinical