Developing a Methodology for the Formation of a System of Attributes of Pathological Vascular Changes in the Fundus

This research aimed to develop a methodology for extracting data from diagnostic images of the fundus blood vessels and methods for their high-precision evaluation, focused on ensuring the diagnostic process standardization, reducing the time of examination and its cost within the framework of evidence-based medicine. Timely and competent diagnosis plays an important role in obtaining an optimal result for treating vascular pathologies. This research evaluated the effectiveness of existing approaches to the analysis of the geometric diagnostic attributes of the fundus vascular system state reflected in the images, which are necessary for identifying pathological vascular changes. A technique was developed for the formation of an optimal system of geometric diagnostic attributes according to the criterion of separability. It was shown that the most effective method for solving this problem is the discriminant analysis method, which, in the presence of a strong connection between certain groups of attributes, makes it possible to decide whether it is expedient to use them and reduce the dimension of the attribute space. Reducing the dimension can significantly reduce the number of calculations. The results of using full-scale images of the fundus with active medical support confirmed the effectiveness of the developed technique.