Authors:
Josef Kohout
and
Martin Cervenka
Affiliation:
Faculty of Applied Sciences, University of West Bohemia, Technická 8, Plzeň, Czech Republic
Keyword(s):
Shape Reconstruction, Point Cloud, Multidimensional Scaling, Muscle Attachments Estimation, Fast Marching, Scalar Distance Field.
Abstract:
Knowledge of muscle attachments on bones is essential for musculoskeletal modelling. A muscle attachment is often represented by points (in 3D) obtained by a manual digitisation system during dissection. Although this representation suffices for many purposes, sophisticated musculoskeletal models commonly require representing a muscle attachment by a surface patch or at least by a closed boundary curve. In this paper, therefore, we propose an approach to automatic shape reconstruction from such point sets. It is based on iso-contour extraction from a scalar field of distances to geodetics connecting the pairs of points (from the input set) as identified by a state-of-the-art algorithm for 2D curve reconstruction running on the input points transformed to 2D. We investigated the performance of 15 existing state-of-the-art algorithms with public implementations on the TLEM 2.0 data set of muscle attachments. The best results were obtained for the lenz algorithm with just one unacceptab
le reconstruction when standard projection onto a best-fit plane was used to transform the input 3D points to 2D. The second algorithm was α-shape with three unacceptable reconstructions, whereas in this case, the multidimensional scaling technique was exploited to transform the points.
(More)