End to End Multitask Joint Learning Model for Osteoporosis Classification in CT Images

Figure 2

Joint framework scheme specific network architecture, including (i) the CT image is sent to YOLOv3 for vertebral positioning; (ii) then, the segmentation module is used to segment the region of interest of the vertebral body, and the feature maps of different scales of the decoding layer are cascaded with the features learned by the ResNet-based convolutional feature extractor, and the key features are obtained by modeling the context features through gated attention; (iii) finally, the features of L1 and L2 are fused by the feature fusion module, and the CT image is classified.