Lung cancer is the deadliest form of cancer in America. To detect the presence of lung cancer, a physician typically performs a bronchoscopy. During a bronchoscopy, the physician attempts to maneuver a bronchoscope through the patient’s airway tree to a suspicious legion identified on a multidetector computed tomography (MDCT) scan. During the procedure, it is difficult to relate the MDCT chest scan to the live bronchoscopic video feed coming from the bronchoscope’s tip. Furthermore, it is difficult for a physician to successfully reach the lesion with the bronchoscope, especially when the lesion is peripherally located.;To this end, research groups have developed bronchoscopy-guidance systems to assist physicians. However, these systems require an attending technician and fail to continuously track the bronchoscope accurately, robustly and in real time. In this thesis, we propose a real-time technician-free bronchoscopy-guidance system that employs continuous tracking and guidance. The system achieves bronchoscope tracking by measuring and interpreting bronchoscopic movements using an external sensor. For guidance, our system uses the bronchoscope’s tracked pose to generate and display real-time virtual bronchoscopy images that are aligned to the live bronchoscopic video. The virtual bronchoscopy images contain guidance information indicating the correct route leading to a region of interest (ROI). Furthermore, we present a colored airway tree indicating the airway-branch diameter to help anticipate difficult bronchoscope maneuvers.;We validate the methods and system strategies on laboratory phantoms derived from human MDCT data as well as video from human bronchoscopy procedures. These experiments demonstrate significant potential for a viable real-time technician-free bronchoscopy guidance system.
展开▼
