This proposal describes an approach to integrate techniques and associated hardware to show the feasibility for automatic analysis of color fundus images. The motivation for the proposed system is a need by ophthalmologists and other physicians for an easy-to-use fundus image capture, image archival, and fundus image analysis system. A quantitative means for analyzing and storing fundus image features such as vascular networks and lesions would aid physicians in assessing the progression of diabetic retinopathy and provide quantitative feedback on the titration regime. The proposed Semi-automatic, Quantitative Analysis of Human Fundus system will evolve into one which will: 1) acquire a fundus image in real-time; 2) digitize the color fundus image; 3) present it to the physician via a graphical workstation environment; 4) archive the image and analyzed features (vascular network and lesions); and 5) provide the physician with an interactive diagnostic tool that allows him/her to revisit same fundus image periodically and quantify changes in features. This system will store unaltered, high resolution images for review. The applications of computer techniques for quantitatively measuring and analyzing retinal images will significantly improve evaluation of patients suffering from diseases such as glaucoma, hypertension, leukemia and diabetic retinopathy.
The successful implementation of this technique has immense commercial potential in the areas of diagnosing and monitoring diabetic retinopathy and hypertension, as well as a number of other medical applications. The potential commercial system would be an end-to-end system, to include a new fundus image acquisition system, a retinal model storage and retrieval system, and a graphical user interface for quantitative computer-assisted fundus analysis for diagnostic support. In Phase II the system will be oriented toward application in the Ophthalmology Department, UNM School of Medicine.
