Object tracking for improved telementoring and telestration

Abstract

Telementoring in image-guided surgery has long since become a worldwide trend, facilitating faster and better surgical expertise distribution, increased quality of learning for beginner surgeons and better clinical outcomes for patients who no longer have to wait for an expert to travel (or do not have to travel themselves). Since the early days of telementoring, telestration (freehand annotation) has been mentioned as an important and valuable part of surgical telementoring. However, to this day, no commercial solutions tailored for surgical telementoring and telestrating that provide telestration stabilization are present. During virtually any surgical telementoring session with telestration, camera movements occur and once the camera moves, the telestrations lose their value unless this movement is calculated and the telestrations’ position is adjusted. Telestration in surgical telementoring needs more attention as an educational and mentoring tool, while at the same time being available without extremely costly equipment and providing all the possible functionalities that a surgeon might need.

The purpose of this work is to estimate the performance of a TLD object tracking algorithm and its openly available alternatives in object tracking / camera motion estimation in laparoscopic surgery videos in context of telementoring and stabilized telestrating, and to propose and architecture and implementation of a prototype solution for stabilized telestration which does not require specialized costly equipment.