Unofficial fork of SOTA lesion segmentation tool nnUNet model, developed by Division of Medical Image Computing, German Cancer Research Center (DKFZ). https://github.com/MIC-DKFZ
This SOTA nnUNet framewok for medical image segmentation is upgraded with Inter-Scanner-Variability-Suppression technique of multi-stage unlearning.
Unlearning represents technique of backpropagating confusion loss, computed from seperate domain predictors on their ability of predicting proper source domain.
Multi-Stage unlearning enables per-downsample-block backpropagation, to enable controlled and uniform confusion loss backpropagation.
Check https://github.com/MIC-DKFZ/nnUNet for nnUNet and framework desciption.