Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Biomedical Engineering

Supervisor

Terry M. Peters

Abstract

Laparoscopic surgery is perhaps the most common minimally invasive procedure for many diseases in the abdomen. Since the laparoscopic camera provides only the surface view of the internal organs, in many procedures, surgeons use laparoscopic ultrasound (LUS) to visualize deep-seated surgical targets. Conventionally, the 2D LUS image is visualized in a display spatially separate from that displays the laparoscopic video. Therefore, reasoning about the geometry of hidden targets requires mentally solving the spatial alignment, and resolving the modality differences, which is cognitively very challenging. Moreover, the mental representation of hidden targets in space acquired through such cognitive medication may be error prone, and cause incorrect actions to be performed.

To remedy this, advanced visualization strategies are required where the US information is visualized in the context of the laparoscopic video. To this end, efficient computational methods are required to accurately align the US image coordinate system with that centred in the camera, and to render the registered image information in the context of the camera such that surgeons perceive the geometry of hidden targets accurately. In this thesis, such a visualization pipeline is described. A novel method to register US images with a camera centric coordinate system is detailed with an experimental investigation into its accuracy bounds. An improved method to blend US information with the surface view is also presented with an experimental investigation into the accuracy of perception of the target locations in space.

chapter2-supplimentary.mp4 (7773 kB)
Supplementary Video for Chapter 2

chapter4-supplimentary.mp4 (3982 kB)
Supplementary Video for Chapter 4

Share

COinS