Research On Segmentation Algorithm Of Lung Cancer PET Images Based On Visual Saliency

As the advanced medical imaging technology, Positron Emission Computed Tomography(PET) images are widely used in clinical diagnosis of heart, nervous system and malignant tumor diseases, especially for the diagnosis, early screening, development stage, and treatment effect of cancer, which is also one of the important means of lung cancer treatment. PET image segmentation algorithm research is of great significance. Because of the characteristics of PET image, the traditional image segmentation algorithm can not obtain ideal segmentation results, and it has some shortcomings in the segmentation speed and segmentation accuracy. In this paper, a Grab Cut image segmentation algorithm based on Itti visual saliency model is proposed, which simplifies the operation steps and improves the efficiency of PET image segmentation and segmentation accuracy, which makes the whole process automation.First, this paper introduces the imaging principle of PET images, and the Itti vision saliency model, further describes the Grab Cut, CA-Grab Cut, Snake three image segmentation algorithm; then, the medical image segmentation algorithm evaluation measure is analyzed, including the reliability, accuracy,regional statistics, efficiency; in addition, a brief introduction to the medical image algorithm research content, for the follow-up study.Secondly, this paper introduces the shortcomings of the traditional Itti model and the Grab Cut algorithm, which leads to the image segmentation algorithm based on significant technology. In this paper, based on the consideration of the PET image as the gray image and the resolution is not high,we propose the improved Itti model and the improved Grab Cut image segmentation algorithm, which solves the problem of the original algorithm in the gray image. At the same time, the operation steps of the users are eliminated,and the operation time of the algorithm is optimized. The operation flow and thesteps of the algorithm are given in detail.Finally, the algorithm is verified by the comparison of lung cancer PET images. The contrast algorithm includes Grab Cut, Snake, CA-Grab Cut, and the comparison is made in the time of partition, the number of user interaction, the overall error rate and the Kappa coefficient. This paper simplifies the operation process. The experimental results show that compared with other algorithms, in the segmentation of lung cancer PET images, this paper has better accuracy,reliability, higher segmentation efficiency, even more simplified user operation.