Segmentation of the Scaphoid Bone in Ultrasound Images

Abstract For the percutaneous fixation of scaphoid fractures, navigated approaches have been proposed to facilitate screw placement. Based on ultrasound imaging, navigation can be carried out in a cost-effective and fast manner, furthermore avoiding harmful radiation. this purpose, efficient architecture for automated segmentation bone volume images is needed. Methods: 2D scaphoid, two architectures are taken into account: nnUNet Deeplabv3+. These trained evaluated newly created dataset consisting 67 annotated in-vivo (4576 slice images). Results: In terms Dice coefficient, achieves 0.67 compared 0.57 distance metrics, shows an average symmetric surface error 0.66mm, while Deeplabv3+ 0.55mm. Conclusion: Fast accurate volumes feasible. Both show competitive results.