Figure 8.

The “mean decrease in Gini coefficient” in random forest models for differentiation of posterior cranial fossa tumors. Separate random forests models were developed for differentiation of all posterior fossa tumors from each other in multiclass analysis, as well as dichotomized classification for the 5 most common posterior fossa neoplasms. The top 10 variables with the highest averaged “mean decrease in Gini coefficient” among from ×100 repeats of 5-fold cross validation are reported. ADC, apparent diffusion coefficient; FLAIR, fluid attenuated inversion recovery.