University of British Columbia
Augmented reality guidance for robot-assisted laparoscopic surgery
Abstract
dc:descriptionThe most common treatment for organ confined prostate cancer is radical prostatectomy (RP), where cancerous prostate is incised out. Nowadays, mostly the da Vinci system is used to do robot assisted laparoscopic radical prostatectomy (RALRP), providing improved dexterity and significantly faster patient recovery times. However similar to open surgery, the RALRP has higher numbers of reported positive surgical margins. This is potentially due to a surgeon’s effort to remove the cancer and preserving the healthy tissue, when the cancerous and non-cancerous boundaries are indistinguishable in the endoscope. Therefore, the objective of this thesis is to clearly display these boundaries and tumors, using Augmented/Mixed Reality (AR/MR) technology, utilizing imaging data such as, magnetic resonance imaging and ultrasound (US). The successful intra-operative AR/MR depends on primarily two things. First, the surgically compatible calibration steps, to map the imaging data correctly in camera image. Second, the visualization of the co-located data to give reliable depth of subsurface structures, which is imperative to patient safety. Unlike existing methods, in our first work, we propose a method that performs the required hand-eye and camera calibrations without using external markers during surgery. To further streamline the process, in another work, we use an optimization scheme to combine both the calibrations in a single step. The method allows to register the robotic data to the camera within minutes. Additionally, we presented an evaluation of a full AR system that registers the phantom US to the camera image. Next, we address the well-known problem of occlusion while visualizing the overlayed imaging data. For this, our deep learning-based method segments surgical instruments in the human RALRPs videos without using any labelled data. In our other two works we explore the color and motion parallax as depth cues to provide a reliable depth judgement. The usefulness of these are validated through user studies showing significantly better depth perception when using our methods. In conclusion, this thesis presents multiple methods to make AR/MR guidance feasible for RALRP by addressing two pressing challenges in the field of surgical AR, i.e. surgically compatible calibrations and reliable visualization of the registered data.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy - PhD
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Electrical and Computer Engineering
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kalia, Megha
- Contributors dc:contributor
-
- Salcudean, Septimiu Edmund
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/84711