A Novel 2D-3D Multi-scale Similarity Registration Method for Bronchoscopy Navigation

Bronchoscope navigation refers to the technology guiding the bronchoscope through the airways. This technology utilizes an imaging guidance system to enable precise navigation of the bronchoscope to the target location, enhancing the accuracy and safety of bronchoscopy procedures. Recent research on bronchoscope navigation has predominantly relied on deep learning, which necessitates high computational capabilities from equipment. This paper introduces a novel multi-scale similarity registration localization algorithm for image-CT, which is built upon the Shape from Shading (SfS) algorithm. This algorithm reconstructs actual depth information from bronchoscope images and utilizes a 3D model derived from airway masks to generate a virtual depth map. It assesses similarity between the two depth maps across three scales: overall depth value similarity within the entire image, regional similarity within the airway area, and cosine and radius similarities of the airway. These similarities from the three scales are combined and weighted to form the objective function for optimizing the position and orientation of the camera. We validated our method using a 3D-printed bronchus phantom, demonstrating its robustness and suitability as a navigation algorithm for bronchoscopy, while also reducing the demand for computational resources.