2020/03/24 by Noura Hamzé, Hamze, Noura, Lukas Nocker +11
Engineering · Medicine · #Medical Imaging and Analysis #Stroke Rehabilitation and Recovery #Musculoskeletal pain and rehabilitation
paper · pdf · doi:10.48550/arxiv.2003.11025
Accurate segmentation of connective soft tissues is still a challenging task,\nwhich hinders the generation of corresponding geometric models for\nbiomechanical computations. Alternatively, one could predict ligament insertion\nsites and then approximate the shapes, based on anatomical knowledge and\nmorphological studies. Here, we describe a corresponding integrated framework\nfor the automatic modelling of human musculoskeletal ligaments. We combine\nstatistical shape modelling with geometric algorithms to automatically identify\ninsertion sites, based on which geometric surface and volume meshes are\ncreated. For demonstrating a clinical use case, the framework has been applied\nto generate models of the interosseous membrane in the forearm. For the\nadoption to the forearm anatomy, ligament insertion sites in the statistical\nmodel were defined according to anatomical predictions following an approach\nproposed in prior work. For evaluation we compared the generated sites, as well\nas the ligament shapes, to data obtained from a cadaveric study, involving five\nforearms with a total of 15 ligaments. Our framework permitted the creation of\n3D models approximating ligaments' shapes with good fidelity. However, we found\nthat the statistical model trained with the state-of-the-art prediction of the\ninsertion sites was not always reliable. Using that model, average mean square\nerrors as well as Hausdorff distances of the meshes increased by more than one\norder of magnitude, as compared to employing the known insertion locations of\nthe cadaveric study. Using the latter an average mean square error of 0.59 mm\nand an average Hausdorff distance of less than 7 mm resulted, for the complete\nset of ligaments. In conclusion, the presented approach for generating ligament\nshapes from insertion points appears to be feasible but the detection of the\ninsertion sites with a SSM is too inaccurate.\n