Research On Ventricle Segmentation And 3D Visualization Based On Magnetic Resonance Images

With the development of medical technology and computer technology,human beings are no longer satisfied with the diagnosis of human diseases through 2D slices,research on medical threedimensional visualization has gradually become a hotspot in medical field,especially in disease diagnosis,medical teaching,surgical navigation etc.,it has great application prospects.This paper mainly studies the implementation process of medical 3D visualization.After the exploration of acquisition and transformation of DICOM images,image segmentation algorithms and inter-layer interpolation algorithms,we design a 3D reconstruction algorithm selection mechanism and make some improvement on algorithm performance,and finally a 3D visualization software is achieved based on MFC framework and OpenGL platform,promoting the deep study of 3D visualization.The specific work are listed as follows:1.An endocardial membrane segmentation algorithm that combined 3D Simplified Pulse Coupled Neural Network(SPCNN)and level set algorithm is proposed after the study of segmentation algorithm of left ventricular and right ventricle MRI images,and comparisons of the characteristics of 2D and 3D segmentation algorithm.The algorithm is tested based on MICCAI2012 database.Considering that there is great difference on file format between medical images and common ones,this paper firstly give the deep analysis of DICOM standard and file format,and then present the file format conversion method of DICOM.2.MC algorithm,PC algorithm and ray casting algorithm are implemented successfully,actually,the main contribution is that the reconstruction process of PC algorithm is improved accordingly.This proposed segmentation algorithm is applied to acquire the endocardial chamber image,and experimental results demonstrated that the sum of the spans after the improvement is smaller than the traditional PC algorithm.We also analyze and compare the advantages and disadvantages of several reconstruction algorithms to achieve the purpose of selecting suitable algorithms for different environments.Besides,a binary inter-layer interpolation algorithm is introduced to achieve the inter-layer interpolation of the segmented contours line,which greatly solve the problem that the inter-layer resolution of the MRI heart image is smaller than the intralayer resolution.By comparing the advantages and disadvantages of various interpolation algorithms,the morphological interpolation is finally selected,and it achieves better results.3.This paper design a 3D visualization software based on MFC framework and OpenGL,and this software implements the functions of DICOM file parsing,image filtering and enhancement,image segmentation,contour line interpolation and 3D reconstruction,it has a simple interface and is easy to operate,which can be used for other 3D visualization studies.