3D long bone reconstruction based on level sets.
In medical imaging a three-dimensional (3D) object must often be reconstructed from serial cross-sections to aid in the comprehension of the object's structure as well as to facilitate its automatic manipulation and analysis. The most popular interpolation scheme for a sequence of image slices is the shape-based method, where object information extracted from a given 3D volume image is used in guiding the interpolation process. The paper presents a level set reformulation of the well-known shape-based method as well as a new automatic level set method, which offers better performance. In particular, we focus on X-ray examinations of long bones, which also requires us to deal with the problem of an optimal slice positioning. To this aim, a 2D version of the proposed algorithm will be used to localize a subset of slices from the entire volume image. A number of experiments were performed on computed tomographic real images to evaluate the proposed approach. The experimental results show a substantial improvement of visual effects (qualitative evaluation) using the proposed method in comparison to both the conventional gray-level interpolation scheme and the shape-based method. Compared with the shape-based interpolation scheme the proposed method has much lower computational cost.