Decision forests in medical imaging analysis

I will talk about recent developments in the area of decision trees and forests and their application to medical image analysis. Decision forests have enjoyed much popularity due to their efficiency and good predictive power. Decision jungles are a variant which uses direct acyclic graphs in place of trees. This yields more compact learner models which generalize better. I will demonstrate such techniques on medical tasks such as segmentation of glioblastoma tumours and localization of spine vertebrae, amongst others.