Алгоритм оценки параметров и прослеживания сосудов на изображениях глазного дна

Abstract

Introduction: Description of retinal blood vessel state by fundus images is a daily procedure in ophthalmological practice. As an object for diagnostic interpretation, digital fundus images are a classic example of medical images of extreme complexity. In order to automate the diagnostic description of retinal blood vessels, we have to highlight the major diagnostic features like the thickness and direction of the vessels. Purpose: We develop an algorithm for estimating the thickness and direction of the fundus vessels, which would allow us to automate the retinal vessel imaging procedures and simplify the monitoring. Results: A technique is proposed for estimating the retinal vessel thickness and direction, based on studying the statistical parameters of fundus images. A normalized difference of the estimates of the universe means was used as observed statistics. It is described how the vessel thickness estimation depends on the test image noise level. An algorithm for tracking retinal vessels is developed. Practical relevance: The proposed algorithm can be used for creating medical expert systems which would allow you to automate the evaluation of retinal vessel state, to monitor the vascular condition, and to assist in diagnostics.