Research Of Medical Image Fusion Based On GPU Acceleration

Different imaging devices have different imaging principle, so the image will beunique. A plurality of different types of medical image fusion processing, can make eachimage advantages complement each other, can make full use of each image information, sothe medical image will be perfect, comprehensive and richer for the clinical diagnosis andtreatment.In both medical research and clinical application, the effect and influence of medicalimage processing technology are growing, it makes us have higher requirements on thespeed of CT and MRI image fusion, it need to rely on the hardware device to get betterperformance for image fusion. Based on this, this article will learn the computer graphicsprocessor (Graphics Processing Unit, the GPU) programming technology, choose the mostpopular GPU development environment-CUDA (Compute Unified Device Architecture)as the platform, to take advantage of its powerful parallel computing features to enhancethe fusion rate. We look forward that, through the use of this technology, clinicalapplications of medical image fusion can meet the requirements of real-time andeffectiveness.The main contents of this paper are as follows:1. This paper briefly introduces the development history and the computationaladvantage of GPU, the basic knowledge and the related knowledge we need to master.Such as the CUDA architecture, programming model, memory model and CUDA programdevelopment environment.2. Based on CT and MRI image as an example, introduces the traditional knowledgeof medical image fusion technology based on CPU, including fusion method, the fusioneffect evaluation index and the fusion process; and based on wavelet transform theory,proposed a new fusion algorithm, the algorithm can bring better results to the fusion ofmedical image.3. Through the comparison of several typical fusion algorithms, choose the wavelettransform method as the algorithm which is the most suitable for using on the CUDAplatform; and introduce the template convolution, analyze the parallelism of Mallatalgorithm; under CUDA platform, achieve the template convolution, Mallat algorithm, andthe fusion of CT image and MRI image. 4. Compared the computing efficiency of two image fusion patterns:GPU parallel-accelerate implementation pattern and the traditional CPU serial-performedpattern, so we can verify that the fusion of medical images based on GPU acceleration hasthe better effect; and analyse the experimental results, summarizes the speeduptrend and the causes of this trend.The paper makes use of the image fusion technology, combined with thecharacteristics of CT image and MRI image, propose a new fusion algorithm for CT andMRI image, is practical to a certain extent; the application of CUDA in medical imagefusion in the field, and the realization of the CT image and MRI image fusion based onGPU acceleration, have innovation to a certain extent; the paper also compared the fusionefficiency in CPU and GPU, the results showed that the execution efficiency of CT andMRI image fusion by CUDA is higher than using CPU, which has advanced to a certainextent.