Automated image analysis, histologic malignancy grading, and survival in patients with astrocytic gliomas. Prognostic significance of nuclear count and nuclear area fraction.

A number of morphometric and densitometric parameters, obtained by automated image analysis, correlate fairly well with classic histologic grading of malignancy in astrocytic gliomas. In this study the predictive value of classic histologic grading of malignancy and of two automatically retrieved nuclear parameters are compared in 199 patients treated surgically for an astrocytic glioma. A 100% follow-up was obtained. Nuclear count and nuclear area fraction correlated fairly well with survival. In patients with 25% of the lowest values of nuclear count and nuclear area fraction 5-year survival was 40%. In patients with 25% of the highest values, less than 5% survived. In contrast, subdivision of the glial tumors into four malignancy grades gave an even better prognostic accuracy. In patients with gliomas of malignancy grade I, about 60% survived for more than 15 years; in malignancy grade IV, no patients survived for more than 5 years.