Research Of Medical Image Segmentation Based On Continuous Max-flow Of CUDA

In recent years,with the demand for automated medical diagnostic technology increasing,medical image processing technology is booming development.Medical imaging technology as the basis for segmentation of medical image processing technology has been widespread concern.One of the most noteworthy is the algorithm accuracy and timeliness.Currently,the accuracy of the medical image segmentation technology has been protected and segmentation algorithm timeliness become an important research direction.In this paper,a GPGPU(General Purpose Graphics Processing Unit)technology to optimize the multiplier based on a continuous flow of the largest medical image segmentation algorithm to improve timeliness.From NVIDIA’s CUDA(Compute Unified Device Architecture)starting to explore its hardware architecture,software architecture and memory architecture works and ways of working as well as the programming model and the advantages of the architecture.Constructed by graph theory and network flow theory s-t network and the establishment of a model of continuous maximum flow,the introduction of pull Lagrangian multiplier to obtain maximum continuous flow algorithm based on multipliers.To enhance the timeliness based Multiplier continuous maximum flow algorithm,this paper presents the maximum continuous flow of medical image segmentation algorithm based on CUDA,the upcoming changes to the original serial segmentation algorithm based on parallel CUDA segmentation.In this paper,given thread allocation algorithm and kernel design,technical solutions,and reduction algorithm to optimize the use of data transmission problems between the host memory and device memory between the end,and further optimize the algorithm processing power.For the proposed method,using four groups of 153 kidneys were images of different resolutions in the MATLAB simulation platform.Experimental results show that under the premise does not affect theaccuracy of segmentation,the use of images of different resolutions for the kidney during the experiment to be segmented image,based on continuous maximum flow algorithm CUDA of timeliness is far superior to the traditional serial continuous maximum flow algorithm.Further,reduction in the use of algorithms improved segmentation algorithm has a better time advantage.