Glaucoma classification based on histogram analysis of diffusion tensor imaging measures in the optic radiation

  • Authors:
  • Ahmed El-Rafei;Tobias Engelhorn;Simone Wärntges;Arnd Dörfler;Joachim Hornegger;Georg Michelson

  • Affiliations:
  • Pattern Recognition Lab, Department of Computer Science and Erlangen Graduate School in Advanced Optical Technologies, Friedrich-Alexander University Erlangen-Nuremberg, Germany;Department of Neuroradiology, Friedrich-Alexander University Erlangen-Nuremberg, Germany;Department of Ophthalmology, Friedrich-Alexander University Erlangen-Nuremberg, Germany;Department of Neuroradiology, Friedrich-Alexander University Erlangen-Nuremberg, Germany;Pattern Recognition Lab, Department of Computer Science, Erlangen Graduate School in Advanced Optical Technologies, Friedrich-Alexander University Erlangen-Nuremberg, Germany;Department of Ophthalmology, Erlangen Graduate School in Advanced Optical Technologies, Interdisciplinary Center of Ophthalmic Preventive Medicine and Imaging, Friedrich-Alexander University Erlan ...

  • Venue:
  • CAIP'11 Proceedings of the 14th international conference on Computer analysis of images and patterns - Volume Part I
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

Glaucoma is associated with axonal degeneration of the optic nerve leading to visual impairment. This impairment can progress to a complete vision loss. The transsynaptic disease spread in glaucoma extends the degeneration process to different parts of the visual pathway. Most of glaucoma diagnosis focuses on the eye analysis, especially in the retina. In this work, we propose a system to classify glaucoma based on visual pathway analysis. The system utilizes diffusion tensor imaging to identify the optic radiation. Diffusion tensor-derived indices describing the underlying fiber structure as well as the main diffusion direction are used to characterize the optic radiation. Features are extracted from the histograms of these parameters in regions of interest defined on the optic radiation. A support vector machine classifier is used to rank the extracted features according to their discrimination ability between glaucoma patients and healthy subjects. The seven highest ranked features are used as inputs to a logistic regression classifier. The system is applied to two age-matched groups of 39 glaucoma subjects and 27 normal controls. The evaluation is performed using a 10-fold cross validation scheme. A classification accuracy of 81.8% is achieved with an area under the ROC curve of 0.85. The performance of the system is competitive to retina based classification systems. However, this work presents a new direction in detecting glaucoma using visual pathway analysis. This analysis is complementary to eye examinations and can result in improvements in glaucoma diagnosis, detection, and treatment.