Approximating Iso-surface Computation Of Boundary Surface Within 3D Biomedical Images

With the rapid development of 3D image technology and its application in the field of biomedical, a lot of 3D biomedical images have been produced. As an important part of visualization, the visualization of 3D biomedical images is driving the development of medical research at the fastest speed. In the visualization technique of 3D images, people usually attempt to extract certain iso-surface from 3D images to represent approximately boundary surfaces of some structures within these 3D images. Considering that some boundary surfaces could be well approximated by certain iso-surface but many boundary surfaces within 3D images actually cannot be well approximated by any iso-surface, it becomes an important problem to distinguish which boundary surface in 3D images could be well approximated by certain iso-surface. In this paper, we will discuss which kind of boundary surface can be approximated by iso-surface and how to compute the optimal iso-surface.At first, we extract the points from the boundary surfaces in 3D images using the algorithm of extracting boundary surfaces. According to the gray value and gradient value of these points from different boundary surfaces, we classify them based on intensity to filter the noise and details that have lower gradient value, while sort them into different classes. Then we extract the gray value of points close to the boundary points; next we detect the points that belong to background and object respectively, and eliminate the influence of partial volume effect through expanding in the direction of gradient or in the opposite direction. At last we analyze the separating feature of the histograms of background and object points, and put forward a rule for judging the error of approximating the boundary surface using the iso-surface, then we bring forward the method that whether the boundary surface can be approximated by certain iso-surface. According to the histograms of background and object points, we apply two algorithms to compute a proper threshold as optimal iso-value: Two-peak and Otsu. For we have known the details of the background and object, we improve the algorithm of Otsu. In the end we validate our method by comparing the iso-line with the contour of object structure in the 2D slices of the 3D images.