Endoscope Navigation and 3D Reconstruction of Oral Cavity by Visual SLAM With Mitigated Data Scarcity

Liang Qiu, Hongliang Ren; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 2197-2204

Abstract


Nowadays computer-assisted surgery (CAS) technologies have been widely used in many aspects of the medical field such as Minimally Invasive Surgery (MIS) or operation focusing on a small surgical site, which has provided significant benefits to patients. However, it is hard for surgeons to determine the accurate poses and surrounding circumstances of the endoscope, due to some restrictions such as narrow field of view (FOV) and misregistration. In this paper, we propose to apply ORBSLAM with a low-cost endoscope to estimate the location of endoscope and create a 3D map for the oral surgery scene, which imposes considerable challenges compared to other human tissue environments, because of the irregular shape, texture-less surface and non-rigid characteristics of the oral cavity. In general, it is very difficult to detect sufficient and effective data for Visual SLAM to realize accurate localization and 3D dense map mainly due to the scarce feature points extracted from tissues and the rare correct matches. In order to reconstruct a denser map for a texture-less oral cavity, laser light markers are used for generating more features, which can mitigate the problem of data scarcity. Besides, we have validated this approach with some experiments on a silicone model of human head. Comparisons between the trajectory/map obtained from ORBSLAM and the ground truth are also provided.

Related Material


[pdf]
[bibtex]
@InProceedings{Qiu_2018_CVPR_Workshops,
author = {Qiu, Liang and Ren, Hongliang},
title = {Endoscope Navigation and 3D Reconstruction of Oral Cavity by Visual SLAM With Mitigated Data Scarcity},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2018}
}