Automated grading of astrocytomas based on histomorphometric analysis of Ki-67 and Feulgen stained paraffin sections. Classification results of neuronal networks and discriminant analysis.

In stereotactically obtained astrocytoma biopsies, four morphometric nuclear parameters were determined with the use of an image analysis system. A special Ki-67 (MIB1)/Feulgen stain made it possible to quantify the essential characteristics of gliomas of the astrocytoma/glioblastoma group: growth pattern, cellularity, proliferation tendency and nucleus pleomorphism. A grading scale based on a cluster analysis resembling the WHO-scheme, which is suitable for automated astrocytoma grading, was developed. Large back propagation neural networks were used and their results compared with those of a classical multivariate discriminant classification analysis. It is possible to show that the neural network technology is superior to the statistical approach for automated astrocytoma grading. Based on the results of our study we believe neural network technology to be useful for tumour grading problems. The presented approach can be generalized for the automated grading of other tumour entities.