简约科技风深度学习网络架构图医学影像分析插画

Stage 2 Input: Add a "PBC Features" block concatenating with CT, Masks, and Coarse Dose. Stage 2 Backbone: Change the standard convolution blocks to "Residual Blocks" (visualize with a curved skip connection arrow on each block). Stage 2 Output: Output 9 channels/heads first. Feed them into a box labeled "Bayesian Fusion". Add a skip line from Stage 1 Output (Coarse Dose) directly into the "Bayesian Fusion" box. Output the final image from the Fusion box. Legend: Add a detailed breakdown of the "Residual Block" at the bottom, similar to the "Dense Block" example provided.