MR imaging of adult supratentorial astrocytomas: an attempt of semi-automatic grading.

Using multiple regression analysis, six MR parameters were correlated with three histological grades among 43 proven adult supratentorial astrocytic gliomas to ascertain important MR parameters and their optimal contributions. Analysis revealed that two parameters, border definition and tumor hemorrhage, were unreliable. Using the remaining four parameters an equation was derived: Tumor grade = 0.32 (ring enhancement) +0.29 (degree of contrast enhancement) +0.13 (heterogeneity) +0.12 (edema) +0.41. Ring enhancement was the most reliable predictor of tumor grade, followed by degree of contrast enhancement. The maximum accuracies of the "semi-automatic" approach using this equation for predicting low-grade astrocytomas, anaplastic astrocytomas, and glioblastoma multiforme were 91%, 83%, and 88%, respectively. Although "semi-automatic" grading provided relatively high accuracy, possible sampling errors and some atypical cases reduced such accuracy.