Open AccessBuilding a model for recognition of morphostructure pathologies in animal tissues
Author(s) -
С. Н. Костарев,
Т. Г. Середа,
Н. А. Татарникова
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1515/5/052005
Subject(s) - pathology , computer science , artificial intelligence , artificial neural network , pattern recognition (psychology) , task (project management) , multilayer perceptron , medicine , engineering , systems engineering
Microscopic images of histological analysis have a complex flat structure. The result of the analysis and interpretation of the diagnosis strongly depends on the qualification of the veterinarian; therefore, the development of methods and models for automated histological diagnosis of diseases is an urgent task. The article considers the known approaches for building expert systems of histological diagnostics. Expert systems of histological analysis can be used in clinical veterinary medicine for advanced training of veterinary pathologists. The possibility of using the theory of finite automata for construction of the device - histological analyzer is shown. As an example of histological analysis the task of recognition of images of morphofunctional changes in rodent tissues on the example of the soft brain of a rat damaged by chlamydia infection is considered. Pathological indicators of morphostructural changes in tissues in case of chlamydia infection have been revealed and explained: erythrocyte slide with clot formation, nucleihyperchrome, tissue edema, hemorrhage and desquamation of endotheliocytes. For image recognition it is suggested to use computer neural network technologies based on multilayer perceptron. The number of synoptic scales for perceptron design has been calculated. Examples are given and construction of training sets on the basis of which the table of truth for neural network training constructed is shown. Computer experiments have shown the possibility of using neural network technologies for recognition of pathology indicators during histological analysis of morphostructural changes in animal tissues.