Figure 2: An overview of the CAiD framework.

Towards learning an optimal embedding space with more discriminative features for medical images, we incorporate a context-aware representation learning with instance discrimination learning. The instance discrimination branch maximizes the (feature-level) similarity between the representations of augmented views x and x′. The context learning branch maximizes the (pixel-level) similarity between original sample s_c and restored ${\widehat{s}}_c$.