Style mixup enhanced disentanglement learning for unsupervised domain adaptation in medical image segmentation

Zhuotong Cai, Jingmin Xin, Chenyu You, Peiwen Shi, Siyuan Dong, Nicha C. Dvornek, Nanning Zheng, James S. Duncan

Abstract: Highlights•Novel framework with data diversification and invariant learning for UDA segmenting.•Disentangled Style Mixup implicitly creates diverse styles in the feature space.•Cascade AdaIN provides more efficient domain assembly.•Dual-level domain-invariant learning enhances underlying domain-invariant mining.•State-of-the-art performance for UDA cardiac and brain segmentation.