Brain Mr Image Segmentation And Visualization

Image segmentation is a vital step in medical image processing and analysis. The diversity andcomplexity of medical images, and the noise introduced in image acquisition, make the precisesegmentation extremely difficult. So far, many segmentation methods have been proposed and tested,yet no common segmentation methods have been found to be suitable for all kinds of images, andoccasionally many algorithms are required to deal with one segmentation task. It is widely acceptedthat the major trends of image segmentation algorithms are speed, accuracy and automation.The medical image visualization technique plays an important role in medical image processingand analysis, and in fact it is one of the most successful applications of visualization in scientificcomputing. To assist diagnosis and therapy, many visualization methods could be used, such as surfacerendering and volume rendering. How to visualize the valuable information contained in medicalimages effectively and intelligibly has been deemed as a major issue in medical volume visualization.With its large data set, the medical volume is born with slow rendering and interactive control speed, soincreasing the rendering speed and improving the accuracy are the primary goals of the relatedalgorithm.The main contributions of this thesis are listed as follows:1. After introducing a variety of image segmentation methods and showing their advantages anddisadvantages, it expounds the principle of watershed transform based on immersion simulation andtells the algorithm implementation procedure. Aiming at the over-segmentation problem in watershedtransform when applied to image segmentation, it explores the marker-based watershed transform andthe method based on region adjacency merging separately.2. Based on brain image gray distribution, it proposes a method of extracting brain marker. Firstly,the thresholds are acquired in terms of gray distribution;Secondly, the gradient image is computed bymeans of Roberts operator;Thirdly, the minimum points between the two thresholds are found;Andlastly, watershed transform on the image is made by using these minimum points above as markers.Experimental results show that this method could separate a brain into different regions and extractbrain as a connected component.3. It describes the classification of medical volume visualization algorithms, and discusses thebasic ideas and characteristics of both surface rendering algorithms and volume rendering algorithms.Then it expounds the algorithm theories, implementation process and experimental results of bothMarching Cubes algorithm and Ray Casting algorithm.4. It discusses the implementation of Ray Casting algorithm in VTK, and describes some speedupschemes. Then it proceeds experiments on some factors influencing the rendering speed and accuracy,and shows the corresponding results.5. It proposes a 3-D surface reconstruction algorithm based on VTK. In the premise that imageshave been classified into different groups, it obtains the gray means of all kinds of tissues based onOTSU algorithm, and renders the iso-surface using that value. Reconstruction results show that it couldtake both the gray distribution of different tissues and differences of imaging equipment into account,and give a smoother and more abundant rendering result than setting iso-value by experience.This thesis mainly consists of six chapters. Chapter 1 presents research value and status of medicalimage segmentation and visualization, and describes main work of this thesis.Chapter 2 gives adiscussion of image segmentation, describes all kinds of segmentation methods and their characteristics,and summarizes present trend of image segmentation. Chapter 3 discusses the basic concepts of digitalimage segmentation, expounds the algorithm theory and implementation process of watershedtransform based on immersion simulation, then describes the solution to the over-segmentationproblem caused by watershed transform, finally presents a new method of extracting brain markers andextracts the whole brain region. Chapter 4 describes the classification of medical image visualizationalgorithm, discusses the theories and characteristics of Marching Cubes algorithm and Ray Castingalgorithm. Chapter 5 discusses Visualization ToolKit and construction procedure of 3-D volume data,analyzes the implementation of Ray Casting algorithm and the factors affecting rendering result, finallypresents a method of 3-D surface construction.Chapter 6 summerizes the main content of this thesis,and points out the study direction aiming at the problems that exists in the current research.