Data Analysis for Comparative Histometry in Pathology - II. Application of an Inductive Knowledge-based System for Classification of Gliomas based on Karyometric Data

Abstract

A knowledge-based system has been used for analysis of morphometric data obtained from 218 neuroepithelial tumors. These tumors have been investigated using standardized morphometric procedures including nuclear area, shape factors, proliferation index Ki-67. It was intended to perform a comprehensive data analysis using different grouping variables (“tumortype” and “tumorgrade”). Additionally, non-discrete numerical variables (“mean nuclear area”, “proliferation index” and “patient's age”) have been tested instead of grouping variables in order to find out, whether different ranges for those variables are exhibited and show significant differences between the cases. Several important informations were provided by the implementation of the knowledge-based system, which are difficult to obtain by means of descriptive data analysis or by means of statistical analysis only. These include for instance a separation of cases with considerably high proliferation index showing a higher age of these patients. Especially, relationships between several variables could be obtained, which could not be detected in conventional statistical data analysis. The information provided by the system has a direct morphological relevance for the evaluation of the tumors. Its use for data analysis of morphometric studies has to be considered as useful, especially when many quantitative variables are under investigation.