Use this URL to cite or link to this record in EThOS:
Title: Intraoperative vision-based instrument tracking and localisation for endovascular procedures
Author: Vandini, Alessandro
ISNI:       0000 0004 7228 9376
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Endovascular procedures are increasingly performed with the benefits of minimal invasiveness, short recovery time and hospital stay, as well as improved patient safety. In addition, several robotic platforms have been recently introduced to overcome the challenges of manually performed endovascular procedures, by improving the tool stability and precision, as well as minimising radiation exposure for the physician and patient. Despite these advances, endovascular procedures are usually performed under 2D image guidance provided by X-ray fluoroscopic images which lack depth perception and soft tissue information. These limitations make challenging and potentially dangerous the navigation of interventional tools through complex anatomies. This research aims to improve the physician's perception during surgery by tracking and localising interventional tools using vision-based approaches. For this purpose, an algorithm that tracks interventional tools in fluoroscopic video sequences is proposed. The algorithm is based on robust features termed SEGlets for segment-like features and their organisation in tracking hypotheses. Furthermore, shape reconstruction of a medical continuum robot is achieved by fusing information extracted from intraoperative images with the kinematics model of the robotic tool. In addition, a 3D vision-based catheter shape reconstruction and localisation algorithm is proposed. The technique does not rely on kinematics modelling of the robot but employs the robot's shape prior and optimal positioning of the imaging system. Finally, an algorithm that addresses simultaneously tracking and shape estimation of a medical continuum robot using a Markov random field framework is proposed. These two steps are usually treated independently in the literature, despite their coupled relationship. Our joint framework, however, proved to be more robust than independently achieving the tracking and shape reconstruction of the robot. Extensive evaluations of the proposed techniques in phantom and surgical data have been performed for demonstrating their potential clinical value towards safe navigation of interventional tools during endovascular procedures.
Supervisor: Yang, Guang-Zhong ; Hamady, Mohamad Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral