The technology of building an expert system for diagnosing malignant nature of invasive tumors of the mammary gland based on a set of quantitative features of the cell nuclei has been developed. Its peculiarity was the presence of weighting coefficients in all the features. Quantitative features were obtained by transforming the initial morphometric data with the help of simple (evaluation of mean values and building of histograms) and complex (regression analysis) mathematical operations. The expert system consisted of one-dimensional X-matrix used for investigations and two-dimensional standard S-matrix. The X-matrix elements were assigned for filling with the quantitative features of the studied sample with a nonestablished diagnosis. The S-matrix elements contained threshold values of quantitative features from the system of diagnostic decision criteria for malignant forms of diseases and their weighting coefficients. Threshold values of nuclear features (larger or smaller) were determined taking into account the range of their values in the groups of malignant and benign pathology. Significance of quantitative features in diagnosing diseases has been assessed. The presence of weighting coefficients allowed diagnosing malignant and benign pathology in a quantitative form by the diagnostic index value. Diagnostic index was calculated by the sum of weighting coefficients of features of the studied sample, which fell within the range of system of the S-matrix diagnostic decision criteria. Clinical trials revealed high efficiency of the developed approach while diagnosis of breast cancer invasive forms at a preoperative stage.
Copyright © 2012 Wiley Periodicals, Inc.