Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Biomedical Engineering


Dr. Terry Peters


Minimally invasive procedures rely on medical imaging instead of the surgeons direct vision. While preoperative images can be used for surgical planning and navigation, once the surgeon arrives at the target site real-time intraoperative imaging is needed. However, acquiring and interpreting these images can be challenging and much of the rich temporal information present in these images is not visible. The goal of this thesis is to improve image guidance for minimally invasive surgery in two main areas. First, by showing how high-quality ultrasound video can be obtained by integrating an ultrasound transducer directly into delivery devices for beating heart valve surgery. Secondly, by extracting hidden temporal information through video processing methods to help the surgeon localize important anatomical structures. Prototypes of delivery tools, with integrated ultrasound imaging, were developed for both transcatheter aortic valve implantation and mitral valve repair. These tools provided an on-site view that shows the tool-tissue interactions during valve repair. Additionally, augmented reality environments were used to add more anatomical context that aids in navigation and in interpreting the on-site video. Other procedures can be improved by extracting hidden temporal information from the intraoperative video. In ultrasound guided epidural injections, dural pulsation provides a cue in finding a clear trajectory to the epidural space. By processing the video using extended Kalman filtering, subtle pulsations were automatically detected and visualized in real-time. A statistical framework for analyzing periodicity was developed based on dynamic linear modelling. In addition to detecting dural pulsation in lumbar spine ultrasound, this approach was used to image tissue perfusion in natural video and generate ventilation maps from free-breathing magnetic resonance imaging. A second statistical method, based on spectral analysis of pixel intensity values, allowed blood flow to be detected directly from high-frequency B-mode ultrasound video. Finally, pulsatile cues in endoscopic video were enhanced through Eulerian video magnification to help localize critical vasculature. This approach shows particular promise in identifying the basilar artery in endoscopic third ventriculostomy and the prostatic artery in nerve-sparing prostatectomy. A real-time implementation was developed which processed full-resolution stereoscopic video on the da Vinci Surgical System.