Fundus Image Analysis for Glaucoma Using Deep Learning and Blood Vessel Information
Brown Bag Lecture by Jongwoo Kim, Ph.D. | 6/12/2018 11:00AM – 12PM | 7th Floor Conference Room, Bldg 38A
Glaucoma is a serious disease caused by damage to the optic nerve, resulting in vision loss. Increased pressure in the eye and/or loss of blood flow to the optic nerve cause nerve fibers to begin to die making the ‘cup’ become larger in comparison to the optic disc. A cup to disc ratio greater than 30% suggests glaucoma. The steps to estimate this ratio are: Region of interest (ROI) detection, optic disc and cup detection, and the ratio estimation. ROI detection is used as a preprocessing step for automatic detection of optic disc and cup areas. This presentation shows an automated method to detect ROI using Convolutional Neural Networks (CNNs). We train two CNNs using fundus images from the MESSIDOR dataset, a public dataset containing 1,200 fundus images. In addition, we estimate blood vessels from the images and use the images embedded with the blood vessels to train two other CNNs. Our method shows good results: The best CNN from the second CNN group (using blood vessel information) indicates over 0.99 accuracy for the MESSIDOR dataset and over 0.97 accuracy for five other image datasets.
Dr. Jongwoo Kim is a staff scientist at the Communications Engineering Branch, Lister Hills National Center for Biomedical Communication. He received a Ph.D. in Computer Engineering and Computer Science from University of Missouri at Columbia, and M.S. and B.S. from Kyungpook National University, Daegu, Korea. His current research interests are biomedical image processing/recognition, machine learning algorithms (deep learning), and biomedical document analysis/processing.