Faculty Advisor or Committee Member

Glenn R. Gaudette, Committee Member

Faculty Advisor or Committee Member

John M. Sullivan, Jr., Committee Member

Faculty Advisor or Committee Member

Gregory S. Fischer, Advisor




"This thesis presents a single-slice based fast stereotactic registration and tracking technique along with a corresponding modular system for guiding robotic mechanism or interventional instrument to perform needle-based interventions under live MRI guidance. The system can provide tracking of full 6 degree-of-freedom (DOF) in stereotactic interventional surgery based upon a single, rapidly acquired cross-sectional image. The whole system is constructed with a modular data transmission software framework and mechanical structure so that it supports remote supervision and manipulation between a 3D Matlab tracking user interface (UI) and an existing MRI robot controller by using the OpenIGTLink network communication protocol. It provides better closed-loop control by implementing a feedback output interface to the MRI-guided robot. A new compact fiducial frame design is presented, and the fiducial is wrapped with a passive resonant coil. The coil resonates at the Larmor frequency for 3T MRI to enhance signal strength and enable for rapid imaging. The fiducial can be attached near the distal end of the robot and coaxially with a needle so as to visualize target tissue and track the surgical tool synchronously. The MRI-compatible design of fiducial frame, robust tracking algorithm and modular interface allow this tracking system to be conveniently used on different robots or devices and in different size of MRI bores. Several iterations of the tracking fiducial and passive resonant coils were constructed and evaluated in a Phillips Achieva 3T MRI. To assess accuracy and robustness of the tracking algorithm, 25 groups of images with different poses were successively scanned along specific sequence in and MRI experiment. The translational RMS error along depth is 0.271mm with standard deviation of 0.277mm for totally 100 samples. The overall angular RMS error is less than 0.426 degree with standard deviation of 0.526 degree for totally 150 samples. The passive resonant coils were shown to significantly increase signal intensity in the fiducial relative to the surroundings and provide for rapid imaging with low flip angles. "


Worcester Polytechnic Institute

Degree Name



Biomedical Engineering

Project Type


Date Accepted





MRI guided intervention, Stereotactic surgery, Image based registration and tracking, Minimally invasive surgery, Passive imaging coil