THE COMBINED USE OF THE DECISION TREE TECHNIQUE AND THE COMPUTER-ASSISTED MICROSCOPE ANALYSIS OF FEULGEN-STAINED NUCLEI AS AN AID FOR ASTROCYTIC TUMOR AGGRESSIVENESS CHARACTERIZATION

A systematic and thus objective method is proposed to characterize ast rocytic tumor aggressiveness. This method relies upon the combined use of a specific decisional algorithm (the decision tree) and 23 paramet ers which include 15 morphonuclear parameters describing the geometric , densitometric, and textural features of a cell nucleus, and 8 parame ters describing the various levels of nuclear DNA content. These 23 pa rameters were objectively quantified by means of the digital cell imag e analysis of Feulgen-stained nuclei. This methodology was used to inv estigate whether it could be applied as a diagnostic tool. The biologi cal model chosen included 12 cell lines adapted to grow in vitro and s temming from 4 astrocytomas (weakly malignant astrocytic tumors) and 6 glioblastomas (highly malignant ones). The 2 additional cell lines we re from two medulloblastomas (MED) (2 highly malignant primitive neuro -ectodermal tumors). The results demonstrate unambiguously that it is actually possible to distinguish between low-grade and high-grade tumo rs on the basis of these parameters, which describe their morphonuclea r features and the amount of their nuclear content. However, a clear-c ut distinction between these different types of tumors can only be att ained when a specific technique is used. In the present case this was the decision tree technique. We were not able to distinguish between t hese various histopathological groups when we used conventional statis tical methods including the one-way-variance analysis of data or the c arrying out of the X(2) test.