A POWERFUL STRATEGY FOR SEGMENTING INTERDIGITATED AND CRIMPED FIBER BUNDLES IN BIOLOGICAL SOFT TISSUES

Biological soft tissues such as tendons, ligaments, articular cartilages and blood vessels consist of collagen fibers that may form collagen fiber bundles (CFBs). In order to develop reliable biomechanical models, it is crucial to have a tool available for the identification of the structural arrangement of CFBs in soft tissue. The task of bundle segmentation becomes quite difficult if the bundles touch each other, show a significant fiber crimp and are only partially visible. A strong fiber crimp leads to high-variance local orientations within a bundle that considerably complicate the bundle segmentation. Straightforward segmentation that combines regions of similar local orientations will fail. In this work we propose a fully automatic vision algorithm that is able to segment even strongly crimped and densely arranged but linearly oriented CFBs