Segmentation And 3D Visualization Of Spine MRI Images

Segmentation and 3D Visualization of medical images are becoming hot topics in medical image processing currently. They are important applications of computer graphics and image processing in biomedical engineering, which have special application value in medical diagnosis and treatment. So research on Segmentation and 3D visualization of medical images has deep significance. This paper centralizes on the segmentation and 3D visualization of spine MRI images. The main work of this paper is as following:Firstly, we perform image pre-processing including image enhancement and interpolation of slice-images for effective segmentation and 3D visualization of spine MRI images. Morphological operations such as opening, closing and top hat are used to enhance the object regions. After comparing and analyzing several commonly used interpolation algorithms, we use the linear gray-based slice-image interpolation method to improve the quantity of 3D visualization.Secondly, we propose a level set method based on spine MRI images. At first, we use a knowledge-based disks detection method to locate and segment the disks, the results of disks segmentation are used for rough segmentation of vertebrae, which are taken as the initialization of level set method, then the RSF-based energy model is used to guide the level set evolution towards vertebra boundaries, at last we get the final segmentation of vertebrae. What's more, we separately construct spine window and vertebrae masks according to the results of disks detection and level set segmentation for the next 3D volume data construction. The spine window and vertebra masks are then used to extract the spine and vertebrae regions from original images.Thirdly, we employ the Ray Casting algorithm to perform 3D reconstruction of the segmented spine and vertebrae. Experiments based on spine MRI images provided by College of Medicine, University of Cincinnati, prove the efficiency of the algorithm, the reconstructed 3D vertebrae have clear 3D profile, and the reconstructed 3D spine could reflect curvature of the spine. In the last of this paper, we implemented the VTK-based 3D visualization of spine and vertebrae under the platform of Visual C++6.0, which has strong practicability, and we can rotate and translate the reconstructed 3D model interactively.