Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.
Non-WSU users: Please talk to your librarian about requesting this thesis through interlibrary loan.
Access Type
WSU Access
Date of Award
January 2023
Degree Type
Thesis
Degree Name
M.S.
Department
Computer Science
First Advisor
Abhilash Pandya
Second Advisor
Marco Brocanelli
Abstract
Positioning the camera during laparoscopic surgery is an essential but often challenging task. The operator will frequently settle for a less than optimal viewpoint due to the mental demand of performing surgery while manually positioning the endoscope or instructing another person to do it. The currently developed autonomous systems are limited in their behaviours or still require some form of operator input. To address this, a modular deep learning based system for autonomous camera control has been developed. Multiple supervised learning models have been trained on 1M+ datapoints from both autonomous and user-operated cameras. The models have an accuracy of 98.6% within 0.001m, although viewpoint optimality is difficult to objectively evaluate. The average response time for the models is 0.017s without considering ROS network latency, and 0.029s including ROS network latency. Additionally, a framework for inverse reinforcement learning (IRL) for autonomous camera control is proposed with a proof of concept.
Recommended Citation
Jawad, Luay, "Autonomous Camera Control For Da Vinci Surgical Systems: A Deep Learning Approach" (2023). Wayne State University Theses. 918.
https://digitalcommons.wayne.edu/oa_theses/918