Marker-Free Registration for Electromagnetic Navigation Bronchoscopy under Respiratory Motion

摘要

Electromagnetic navigation bronchoscopy requires the accurate registration of a preinterventional computed tomography (CT) image to the coordinate system of the electromagnetic tracking system. Current state-of-the-art registration methods are manual or do not explicitly take patient's respiratory motion and exact airway shape into account, leading to relatively low accuracy. This paper presents an automated registration method addressing these issues. Electromagnetic tracking data recorded during bronchoscopic examination is matched to the airways by an optimizer utilizing the Euclidean distance map to the centerline of the airways for automated registration. Using a cutaneous sensor on the chest of the patient allows us to approximate respiratory motion by a linear deformation model and adopt the registration result in real time to the current respiratory phase. A thorough in silico evaluation on real patient data including CT images taken in 10 respiratory phases shows the significant registration error decrease of our method compared to the current state of the art, reducing the error from 3.5 mm to 2.8 mm.