Master of Engineering Science
Dr. Terry Peters and Dr. Ali Khan
Modern neurosurgical procedures often rely on computer-assisted real-time guidance using multiple medical imaging modalities. State-of-the-art commercial products enable the fusion of pre-operative with intra-operative images (e.g., magnetic resonance [MR] with ultrasound [US] images), as well as the on-screen visualization of procedures in progress. In so doing, US images can be employed as a template to which pre-operative images can be registered, to correct for anatomical changes, to provide live-image feedback, and consequently to improve confidence when making resection margin decisions near eloquent regions during tumour surgery.
In spite of the potential for tracked ultrasound to improve many neurosurgical procedures, it is not widely used. State-of-the-art systems are handicapped by optical tracking’s need for consistent line-of-sight, keeping tracked rigid bodies clean and rigidly fixed, and requiring a calibration workflow. The goal of this work is to improve the value offered by co-registered ultrasound images without the workflow drawbacks of conventional systems. The novel work in this thesis includes:
- the exploration and development of a GPU-enabled 2D-3D multi-modal registration algorithm based on the existing LC2 metric; and
- the use of this registration algorithm in the context of a sensor and image-fusion algorithm.
The work presented here is a motivating step in a vision towards a heterogeneous tracking framework for image-guided interventions where the knowledge from intraoperative imaging, pre-operative imaging, and (potentially disjoint) wireless sensors in the surgical field are seamlessly integrated for the benefit of the surgeon. The technology described in this thesis, inspired by advances in robot localization demonstrate how inaccurate pose data from disjoint sources can produce a localization system greater than the sum of its parts.
Pardasani, Utsav, "Neurosurgical Ultrasound Pose Estimation Using Image-Based Registration and Sensor Fusion - A Feasibility Study" (2016). Electronic Thesis and Dissertation Repository. 4194.