Skeleton Graph-Based Ultrasound-CT Non-Rigid Registration

Autonomous ultrasound (US) scanning has attracted increased attention, and it been seen as a potential solution to overcome the limitations of conventional US examinations, such inter-operator variations. However, is still challenging autonomously accurately transfer planned scan trajectory on generic atlas current setup for different patients, particularly thorax applications with limited acoustic windows. To address this challenge, we proposed skeleton graph-based non-rigid registration adapt patient-specific properties using subcutaneous bone surface features rather than skin surface. end, self-organization mapping successively used twice unify input point cloud extract key points, respectively. Afterward, minimal spanning tree employed generate graph connect all extracted points. appropriately characterize rib cartilage outline match source target cloud, path from optimized by maximally maintaining continuity throughout each rib. validate approach, manually one volunteer seven CT clouds patients. The results demonstrate that more effective robust in adapting inter-patient variations ICP (distance error mean $\pm$ SD: notation="LaTeX">$5.0\pm \rm{1.9}~mm$ vs notation="LaTeX">$8.6\pm \rm{6.7}~mm$ CTs).