Tissue Shape Acquisition with a Hybrid Structured Light and Photometric Stereo Endoscopic System.

In situ 3D reconstruction from endoscopic images is important to determine the correct course of action for, e.g., treatment of abnormal growths. Currently, the endoscopist has to rely solely on visual cues in order to infer the growth's shape and size and determine an appropriate treatment. However, tissue uniformity and scale ambiguity from traditional monocular endoscopes make this visual assessment prone to errors and time consuming. We propose a practical system to densely reconstruct both shape and size of tissues with minimal modifications to a standard endoscope. We present a custom single-fiber structured light probe projecting a wave pattern on the tissue surface that allows semi-dense reconstruction with few ambiguities. Based on the coarse reconstruction, we retrieve the surface reflectance parameters according to a hybrid diffuse/specular model which are used to initialise a close-range Photometric Stereo reconstruction. By taking into account the tissue characteristics and the light fall-off, our Photometric Stereo formulation provides dense metric 3D shape information without the need for surface normal integration. A preliminary study was carried out both on phantoms and ex vivo samples of human tissue.