| Abstract [eng] |
Summary Analysis of 3D Medical Images Nowadays medicine generates twice as much medical images as it generated several years ago. Manual analysis and segmentation of this images is a very difficult task which is time consuming and also has quite high error rate. Computer-aided diagnosis by using computer algorithms tries to help medical personnel to make image analysis fast and in a high quality. In this thesis computer tomography process will be deeply discussed. Some principles will be discussed about how CT works and what are main parameters which influences medical images. Format DICOM will be also discussed to show how does are medical images stored and transported via the web. Image binarization techniques will be presented. Several algorithms to calculate threshold values will be discussed along with their mathematical explanation. In practical part of this thesis these methods will be implemented and tested with real CT images. Then best method will be chosen to make whole chest bones segmentation. After final segmentation bones will be visualized in 3D. In the second part of this thesis solution for lung segmentation will be presented. Firstly, algorithms for image processing will be discussed. Image labeling techniques, mathematical morphology, background removal techniques will be outlined. After that method to segment lung will be presented. After successful lung segmentation marching cubes algorithm for surface reconstruction will be introduced. Finally 3D visualization solution will be presented. We will briefly cover WebGL and Three.js library usage. Practical part of lung segmentation and reconstruction is also presented in this thesis. Implementation consists of three parts: lung points segmentation, surface reconstruction and visualization. First two parts are implemented using Java programming language, third part is implemented in Java Script and HTML. |
