Faculty Advisor

Prof. Michael Gennert

Faculty Advisor

Prof. Eduardo Torres-Jara

Faculty Advisor

Prof. Gregory Fischer




This thesis deals with automatic localization and tracking of surgical tools such as needles in Magnetic Resonance Imaging(MRI). The accurate and precise localization of needles is very important for medical interventions such as biopsy, brachytherapy, anaesthesia and many other needle based percutaneous interventions. Needle tracking has to be really precise, because the target may reside adjacent to organs which are sensitive to injury. More over during the needle insertion, Magnetic Resonance Imaging(MRI) scan plane must be aligned such that needle is in the field of view (FOV) for surgeon. Many approaches were proposed for needle tracking and automatic MRI scan plane control over last decade that use external markers, but they are not able to account for possible needle bending. Significant amount of work has already been done by using the image based approaches for needle tracking in Image Guided Therapy (IGT) but the existing approaches for surgical robots under MRI guidance are purely based on imaging information; they are missing the important fact that, a lot of important information (for example, depth of insertion, entry point and angle of insertion) is available from the kinematic model of the robot. The existing approaches are also not considering the fact that the needle insertion results in a time sequence of images. So the information about needle positions from the images seen so far can be used to make an approximate estimate about the needle position in the subsequent images. During the course of this thesis we have investigated an image based approach for needle tracking in real-time MR images that leverages additional information available from robot's kinematics model, supplementing the acquired images. The proposed approach uses Standard Hough Transform(SHT) for needle detection in 2D MR image and uses Kalman Filter for tracking the needle over the sequence of images. We have demonstrated experimental validation of the method on Real MRI data using gel phantom and artificially created test images. The results proved that the proposed method can track the needle tip position with root mean squared error of 1.5 mm for straight needle and 2.5mm for curved needle.


Worcester Polytechnic Institute

Degree Name



Computer Science

Project Type


Date Accepted





Needle Tracking, Kalman Filter, Hough Transform