CT images inpainting based on graph-cut segmentation

Abstract

Damaged pelvic bone repair requires 3D printing of the prosthesis. Traditional medical methods require high time and labor input. At present, the use of computer-assisted preoperative planning can largely assist doctors in quickly determining the cause of the disease and giving timely advice. Appropriate treatment with the patient. In view of the fact that traditional preoperative planning methods rely solely on the physician's personal experience, which can easily lead to errors in judgment and fail to give the patient correct and efficient treatment opinions,this article introduces an algorithm for automatically segmenting human bones and a method of CT images inpainting. The automatic segmentation algorithm uses the segmented CT images to input the GAN model for image inpainting, and then performs the 3D reconstruction of the marching cubes on the repaired CT images. The results of the experiment present that the human bone automatic segmentation is far superior to traditional segmentation algorithms in RMSD and MSD indicators, and the CT images inpainting indicators after GAN model repair are also better than traditional WGAN and DCGAN models.