| Popis: |
The methods of the multivariate mathematics have been applied in several studies to increase the diagnostic reliability of medical decision support system. In the recent years some new algorithms for decision support (fuzzy logic) and for pattern recognition (neural nets), both specified by nonlinearity, were developed. This paper provides results for the application of this methods in the area of quantitative cytology and the comparison with the traditional classifiers. 21 normal, 15 dysplastic, 23 malignant, Feulgen stained gastric imprint smears were analysed on a Leitz Miamed DNA equipment. The determination of mean DNA content, the 2c deviation index (2cDI), 5c Exceeding rate (RcER), G1,S,G2 phase fraction ratios, cell nucleus area, form factor was performed. The discriminant analysis classified correctly the 95.6% of malignant cases, 86.7% of dysplasias, and 80.7% normal cases. Our diagnostic system using fuzzy logic made the diagnostic borders fine tuneable, and reliable. The back propagation neural net could classify all three diagnostic groups above 95% correctly. The application of nonlinear computational methods made the diagnostic system more reliable. The application of these algorithms are encouraged. |
