Karas O. Videopolarimetric system for analysis of blood plasma films images in the assessment of mammary glands pathologies.

The object of the dissertation research is the processes of measurement and analysis of polarization and anisotropic parameters of human plasma films with subsequent diagnosis based on their pathological changes in the mammary glands. The aim of the dissertation research is to increase the reliability of diagnosing two-dimensional laser polarimetry to assess pathological changes in the mammary glands by using methods of polarization Jones-matrix mapping of human plasma films in combination with statistical and correlation analysis of images and differentiation based on neural network technologies. During the dissertation the following research methods were used: methods of applied optics for analysis of polarization of a field scattered by a biological object when irradiated with laser polarization radiation, methods of system analysis, methods of statistical analysis of two-dimensional distributions of random polarization parameters, methods of artificial neural networks to determine the structure and parameters of neural networks, methods of neural network technologies for estimating pathological changes of biological objects, error matrix method for estimating classification. Theoretical results: for the first time an information model of decision support in assessing the state of the mammary glands by Jones-matrix mapping of blood plasma films using statistical and correlation analysis of images to form diagnostic features and differentiate pathologies, which minimized uncertainty in assessing such changes; for the first time, relationships were found between a set of statistical and correlation moments that characterize the coordinate distributions of Jones-matrix images of human blood plasma films and physiological states of "normal", "fibroadenoma" of the mammary glands; The method of Jones matrix mapping of biological layers was improved, in which the obtained Jones matrix images of human plasma films are subjected to statistical and correlation analysis with further differentiation based on neural network technologies, which increased the reliability of diagnosis to 93%. The practical results are to improve the videopolarimetric system for the analysis of images of blood plasma films in combination with their statistical and correlation analysis, which allowed to determine the principles of improving polarization systems in various fields of medical practice; developed algorithmic support for a videopolarimetric system for the analysis of images of blood plasma films, which can be used in screening studies of breast pathologies for early detection of patients at increased risk of cancer or cancer; and also developed a database of informative features containing the ranges of statistical and correlation moments of maps of imaginary and real elements of Jones matrices of blood plasma films, which correspond to physiological states "norm", "fibroadenoma" of mammary glands, which allowed for automated diagnostic process. The results obtained in the dissertation are recommended for implementation at Vinnytsia National Medical University named after M. I. Pyrogov and in the private enterprise "Photonics Plus".