DrasCLR: A self-supervised framework of learning disease-related and anatomy-specific representation for 3D lung CT images

Ke Yu, Li Sun, Junxiang Chen, Maxwell Reynolds, Tigmanshu Chaudhary, Kayhan Batmanghelich

Abstract: Highlights•Self-supervised method extracting disease-specific features from 3D medical data.•Two domain-specific contrastive learning cues leverage anatomical similarities.•Novel 3D convolutional layer with anatomical location-dependent kernels.•Pretrained model generalizes well across downstream prediction tasks.•Improved label-efficiency in lung CT image segmentation.

External IDs:dblp:journals/mia/YuSCRCB24