advertisement
Abstract #19663 Published in IGR 9-4
Automated interpretation of optic nerve images: a data mining framework for glaucoma diagnostic support
Abidi SS; Artes PH; Yun S; Yu JMedinfo 2007; 12: 1309-1313
Confocal Scanning Laser Tomography (CSLT) techniques capture high-quality images of the optic disc (the retinal region where the optic nerve exits the eye) that are used in the diagnosis and monitoring of glaucoma. We present a hybrid framework, combining image processing and data mining methods, to support the interpretation of CSLT optic nerve images. Our framework features (a) Zernike moment methods to derive shape information from optic disc images; (b) classification of optic disc images, based on shape information, to distinguish between healthy and glaucomatous optic discs. We apply Multi Layer Perceptrons, Support Vector Machines and Bayesian Networks for feature sub-set selection and image classification; and (c) clustering of optic disc images, based on shape information, using Self-Organizing Maps to visualize sub-types of glaucomatous optic disc damage. Our framework offers an automated and objective analysis of optic nerve images that can potentially support both diagnosis and monitoring of glaucoma.
Dr. S.S. Abidi, NICHE Research Group, Faculty of Computer Science, Department of Ophthalmology and Visual Sciences, Dalhousie University, Halifax, Canada. sraza@cs.dal.ca
Classification:
6.9.1.1 Confocal Scanning Laser Ophthalmoscopy (Part of: 6 Clinical examination methods > 6.9 Computerized image analysis > 6.9.1 Laser scanning)