In order to improve the accuracy and efficiency of medical image segmentation, this paper designs and proposes a medical image visualization method containing Sobel edge detection operator and 3D UNet network based on deep learning and edge detection. The 3D U-Net network is used to capture the morphological and edge features of medical images on the public dataset, and the image binarization is performed on the result of its operation. The binarized image processed by corrosion and expansion algorithms is multiplied by the corresponding elements of the matrix with the medical image to obtain the visualization of the medical image. Different comparison algorithms and data sets are selected to verify the effectiveness of the optimized 3D U-Net network module and feature fusion module. Parameter settings are carried out, and the LIDC-IDRI dataset is used as the algorithm training base data to analyze the segmentation accuracy of the image processing method that fuses the edge detection operator with the 3D U-Net network. The algorithm ablation experiments are carried out according to different pruning degrees and training methods. The algorithm in this paper can achieve more than 80% segmentation accuracy on LIDC-IDRI dataset, in which the segmentation accuracy of liver reaches 97.1%.