Joint Dense Residual and Recurrent Attention Network for DCE-MRI Breast Tumor Segmentation

Figure 7

Recurrent attention module. ‘Conv 3×3’ represents a convolution block composed of a 3×3 kernel size convolution, GN, and PReLU. ‘GRU’ denotes a gated recurrent unit. ‘’ is the initial hidden state, and ‘’ and ‘’ are the output hidden states of GRU. ‘GMP’ indicates global max-pooling. See Section 2.5.3 in Section 2.5 of the main text for further explanation.