Research on spinal orthopedic locator image segmentation algorithm in X-ray images

In the process of orthopedic 2D/3D medical image registration, a spinal orthopedic locator is needed to assist registration. There is a spinal orthopedic locator image in the orthopedic X-ray image that hinders registration. It is necessary to segment and remove the spinal orthopedic locator image in the X-ray image to improve registration. Accuracy. In response to the above problems, this paper proposes the SR-UNet deep learning algorithm model for segmentation of spinal orthopedic locator images in orthopedic X-ray images. Based on the advantages of U-shaped network in medical image segmentation, SR-UNet network combines U-shaped network, The characteristics of residual network and asymmetric network can accurately segment the spinal orthopedic locator image. The SR-UNet network model is trained through the self-built training set and verification set, and the performance of the simulation with accurate segmentation is evaluated through the Dice Similarity Coefficient (DSC) index. The SR-UNet deep learning algorithm model ensures high-precision segmentation effects and can accurately segment spinal orthopedic locator images in orthopedic X-ray images. In the case of the same image data set, the DSC index exceeds other model networks.