Premium
Technology of creation of an expert system for diagnosing thyroid pathology based on a set of qualitative signs of cell atypia
Author(s) -
Kirillov Vladimir,
Gladyshev Alexander,
Demidchik Evgeny
Publication year - 2010
Publication title -
microscopy research and technique
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.536
H-Index - 118
eISSN - 1097-0029
pISSN - 1059-910X
DOI - 10.1002/jemt.20853
Subject(s) - atypia , nuclear atypia , matrix (chemical analysis) , weighting , pathology , set (abstract data type) , feature (linguistics) , computer science , medicine , mathematics , artificial intelligence , radiology , chemistry , immunohistochemistry , chromatography , programming language , linguistics , philosophy
Abstract An expert system for differential diagnosis of thyroid pathology has been developed, in which the function of transforming qualitative signs of cell atypia into the quantitative form is realized. It based on the set of qualitative signs of cell atypia and works in the mode by the question—answer principle. An expert system (software) contained the X‐matrix and six standard S‐matrices. The X‐matrix was designed for filling with the alternative answers yes/no regarding the presence of cells with characteristic signs of atypia in the visual fields of a preparation. S‐matrices characterized the main forms of thyroid diseases, such as papillary and follicular cancer, follicular adenoma, autoimmune thyroiditis, nodular colloid and diffuse toxic goiter, by their set of yes/no answers. The answers entered to the X‐matrix were compared with answers of the S‐matrix. Each element of each S‐matrix, besides the answers yes/no, contained its weighting coefficient corresponding to a qualitative feature of a certain nosologic form. Weighting coefficient of each feature for every nosologic form was calculated by their frequency of occurrence in 30 visual fields. Comparison of X‐matrices with S‐matrices and verification of diagnosis according to the value of a diagnostic index with a percentage probability was performed in the automatic mode. Diagnostic index was determined by the sum of weighting coefficients of features with the coincident yes/no answers, while comparing X‐matrix elements with the corresponding S‐matrix elements. The software displayed on the monitor is the list of nosologic forms in the order of decreasing the diagnostic index values. Microsc. Res. Tech. 73:1091–1100, 2010. © 2010 Wiley‐Liss, Inc.