Research On 3D Medical Image Segmentation Algorithm Based On Surfacelet Transform

With the development of medical imaging technology and the improvement of computer computing capabilities,3D medical image segmentation has become an important research direction.However,limited by the principles of medical imaging and the complexity of human organs and structural shapes,3D medical image segmentation still faces enormous challenges.The traditional 3D medical image segmentation is based on 2D image segmentation and then reconstructed,due to the inability to use the structural information between adjacent frames,the segmentation is inaccurate and time-consuming.The existing segmentation techniques directly performed in 3D space,such as computer assisted visualization and analysis software system(CAVASS),statistical shape model,and graph cut theory.Due to various types of medical images,there is a problem of insufficient generalization ability in different tools or algorithms.Thus,we present a 3D medical image segmentation method based on surfacelet transform.This article mainly includes the following research contents:(1)This paper studied the basic theoretical knowledge of surfacelet transformation and analyzed the concept of directional filter.The distribution characteristics of the subband coefficients decomposed by the surfacelet transform are introduced in detail.Through experimental analysis,after the surfacelet transform of the 3D medical signal,the boundary region of the image is distributed in the high-frequency subband coefficients,and the main detail information is distributed in the low-frequency subband coefficients.The segmentation quality of the 3D image is enhanced by processing the high and low frequency coefficients of the surfacelet transform domain.(2)This paper proposed a 3D medical image segmentation algorithm based on surfacelet transform and canny operator in view of the structural characteristics of 3D medical images.Surfacelet transform has the advantages of translation invariance,perfect reconstruction,and low redundancy.It can handle multidimensional signals parallel and decompose the signals in multi scale and multi direction.First,the 3D image is decomposed into three dimensional surfacelet transform domain.Adaptive filtering of the high frequency subband coefficient and normalized operation of the low frequency subband coefficient based on the distribution characteristics in the surfacelet transform domain.Second,surfacelet reconstruction based on the treated coefficient.Third,the canny operator is used to extract the edges of high and low frequency images,and a fusion operation is implemented to obtain whole edges with high accuracy.The experimental results show that the proposed method can separate effectively signal from the noise,enhance the weak edges and keep stable sparse coefficients by employing different processing strategies for the high and low frequency subband coefficients of the surfacelet transform.The method can obtain accurate edges and full contours of the interested objects in 3D medical image.