A high accuracy multi-image registration method for tracking MRI-guided robots

Abstract

Recent studies have demonstrated an increasing number of functional surgical robots and other devices operating in the Magnetic Resonance Imaging (MRI) environment. Calibration and tracking of the robotic device is essential during such MRI-guided procedures. A fiducial tracking module is placed on the base or the end effector of the robot to localize it within the scanner, and thus the patient coordinate system. The fiducial frame represents a Z shape and is made of seven tubes filled with high contrast fluid. The frame is highlighted in the MR images and is used in localization. Compared to the former single image registration method, multiple images are used in this algorithm to calculate the position and orientation of the frame, and thus the robot. By using multiple images together, measurement error is reduced and the rigid requirement of slow to acquire high quality of images is not required. Accuracy and performance were evaluated in experiments which were operated with a Philips 3T MRI scanner. Presented is an accuracy comparison of the new method with varied number of images, and a comparison to more traditional single image registration techniques.