Fig. 5 | Scientific Reports

Fig. 5

From: Automatic detecting multiple bone metastases in breast cancer using deep learning based on low-resolution bone scan images

Fig. 5

(a) Pixel-level mask and entropy. (b) Architecture of ST_Encoder. (MLP = multi-layer perceptron, LN = LayerNorm). W-MSA and SW-MSA denote multi-head self-attention modules with regular and shifted windowing configurations, respectively. (c) Architecture of T_Decoder. (MHA = multi-head attention, FFN = feed forward network). This block treats entropy as an object query. After converting entropy into a suitable vector format, T_Decoder feeds the processed query into the MHA block together with the origin features as key and value to perform cross-attention.

Back to article page