Surface reconstruction of biological organs from serial microscopic sections
Roman Ďurikovič, Kazufumi Kaneda, Hideo Yamashita · 2002
An effective system for 3-D reconstruction of objects from a set of planar cross-sections is presented in this paper. First, the noisy intensity images are aligned, and consequently, all images are segmented sequentially from top to bottom. To extract contours from a series of cross-sectional images, a new semi-automatic method based on an active contour model is developed. The active contour model uses the texture information stored in texture feature vectors for each small area of tissue. For the dynamic iteration process, we can use the final contours as the initial contour in the next section, assuming that the difference between consecutive cross-sections is small. In this way, the obtained planar contours are further processed to find the topological correspondences. This step is promoted by criteria using geometrical and topological information. Our system was successfully applied to the contour extraction and reconstruction processes of a sequence of 636 microscopic images (7 /spl mu/m in thickness), as demonstrated in our video animation.