Numerical grading of astrocytomas.

Ninety-three astrocytomas from biopsy material including glioblastomas ('astrocytomas grade 4') were graded according to Kernohan (1949) by light microscopy. Feulgen sections were subjected to an automated microscopic analysis to obtain morphometric-densitometric data of the tumour cell nuclei. These quantitative and reproducible data showed a significant correlation with malignancy expressed in terms of Kernohan and prove automated image analysis to be a valuable tool in the grading of gliomas. This is in particular true if the nuclear parameters determined by image analysis are completed by histologic features which were recorded semiquantitatively by subjective light microscopic evaluation as is usual in clinic pathologic diagnosis of brain tumours. A quadratic discriminant analysis of morphometric-densitometric data of tumour cell nuclei and semiquantitative microscopic data gave a 94% agreement with subjective grading.