Osteolysis: A Literature Review of Basic Science and Potential Computer-Based Image Processing Detection Methods

Figure 12

The multilevel Seg-UNet model with patch- and global-based approaches (top) and the block demonstration of the model (bottom). In the block diagram, E represents the encoding block; D depicts the decoding block; the three branch blocks H_seg, H_dist, and H_clas denote multitask learning [78].