Recently we have a new paper accepted by MICCAI 2016: Dynamically Balanced Online Random Forests for Interactive Segmentation. The work is part of the Gift-Surg project.
We proposed a new type of Random Forest aimed to deal with imbalanced training data with changing imbalance ratio, which occurs in the context of scribble-based interactive segmentation. The proposed method has a better performance in dealing with data imbalance problem in online learning than traditional Online Random Forests. We have applied it to placenta segmentation in this paper.
We have made the code available online.
Authors: Guotai Wang, Maria A. Zuluaga, Rosalind Pratt, Michael Aertsen, Anna L. David, Jan Deprest, Tom Vercauteren, Sebastien Ourselin.