Shehna K. Support system for making diagnostic decisions in the analysis of digital medical images.

The dissertation is devoted to solving the current scientific and practical problem of increasing the efficiency of detecting anatomical and pathological structures on low-contrast halftone images of the breast by developing a formalized model of the mammographic examination process, mathematical methods for implementing its individual stages, and developing a decision support system using modern information technologies. The aim of the dissertation is to improve methods of analysis of halftone biomedical images in mammographic decision support systems. The object of research is the process of analysis anatomical and pathological structures on digital mammograms. The subject of research methods of selection and classification of anatomical and pathological structures on digital mammograms are presented . According to the results of the research, the following scientific results were obtained: – the general model of the mammographic examination process, which includes functional, structural, informational and mathematical models, has been improved by formalizing the stages : input of diagnostic data and preliminary processing of mammograms taking into account the peculiarities of their display on low-contrast halftone images of the mammary gland; fractal processing of digital mammograms for the purpose of forming a system of diagnostic signs; development of a diagnostic decisive rule for the formation of a computer diagnosis; development of a decision regarding the diagnostic and treatment process , which allowed to apply formal methods of implementation of the noted stages; – the method of identifying and classifying anatomical and pathological structures on digital mammograms, including intraductal formations and microcalcifications by calculating their fractal dimension, was further developed, which made it possible to form a vector of diagnostic signs for making a computer diagnosis; – for the first time, a combined diagnostic decisive rule was developed based on the modification of the method of comparison with the standard by introducing expert information on the structure of symptom complexes when calculating the coordinates of class prototypes, which allows taking into account both the objective probabilistic component and the subjective component of the diagnosis process, which is a formalization expert assessment of the structure of the symptom complex of the specified disease. Variants of joint use of its components (collective decision rules, weighting and summarization of evaluations) are offered. The practical significance of the obtained results for the field of information technologies is that the methods of fractal processing of digital mammograms developed in the dissertation serve as a scientific and methodological basis for the development of appropriate information, algorithmic and software. A computer system has been developed that provides support for decision-making in the classification of digital mammograms for the presence/absence of pathological structures for the purpose of further diagnosis. The results of the dissertation work were implemented in the educational process of KhPI National Technical University at the Department of Computer Engineering and Programming in the study of the disciplines "Signal and Image Processing", "Design of Computer Diagnostic Systems", "Fundamentals of Scientific Research" and in diploma and coursework design.