Endoscopic view expansion for tracheal intubation using feature-based image-sequence stitching

Field-of-view is crucial in flexible endoscopic intubation because visual information is the primary source for surgical decision making. In the narrow upper airway, endoscopic view expansion (EVE) can potentially improve doctors’ perception of the environment and facilitate tracheal intubation. This study investigated the feasibility of upper airway EVE using feature-based image sequence stitching. We designed a strategy for sparse candidate frame selection based on structure similarity index measure and constructed a dataset from intubation endoscopic videos of 70 adult patients. An AFFINE based SIFT approach was adopted to overcome the influence of tissue deformation in the upper airway endoscopic recordings. The approach was combined with the moving direct linear transformation for optimal local alignment. Finally, an incremental stitching scheme was proposed for real-time EVE application, and verifications showed an average field expansion rate of 148 % over ten frames near key anatomical structures. This study demonstrate the feasibility of flexible endoscopic view expansion and lays the foundation for further applications in endotracheal intubation scenarios. © 2024