Supplementary Figure 2 from Rapid Intraoperative Diagnosis of Pediatric Brain Tumors Using Stimulated Raman Histology

Random forest model 2. A) Example decision tree from random forest model 2. B) Out-of-bag, low-grade, and high-grade tumor classification error across 500 grown decision trees. C) Image feature importance for model 2. Similar to model 1, nuclei density and TAM density are the most important predictors; however, nuclear morphology parameters and detection of anaplasia play a more important predictive role for classifying low-grade and high-grade tumors.