Fig. 4From: Automatic segmentation model of intercondylar fossa based on deep learning: a novel and effective assessment method for the notch volumeA Structure of the residual block. Each residual block (RB) consists of two batch normalization layers, two rectified linear unit(ReLU)layers and two 3 × 3 convolutional layers. B The structure of the combine block. Each combine block consists of two inputs, a rectified linear unit(ReLU)layer and a convolutional layerBack to article page