Addressing Shading-Based Laparoscopic Registration

Abstract

Surgical augmentation is addressed with a novel approach driven by the shading information captured by conventional endoscopes. The method is applicable even for organs or surfaces with no salient features, where the matching problem in motion algorithms is prone to fail. The idea is to produce 3D reconstructions of organs from intraoperative video-endoscopic images, and subsequently align them to the preoperative volumetric models. A propagating shape from shading (SfS) technique is adapted to the perspective camera model with a point source of light located at the camera centre. The alignment is performed based on an iterative closest point algorithm with two additional steps, one to update the scale of the perspective reconstruction and the other to tune the albedo and light intensity effects in the SfS reconstruction. Results from both synthetic and real endoscopic images show good spatial alignment of the shapes. We therefore believe shading-driven augmentation is a feasible approach for applications like laparoscopic liver hepatectomy.