BOSTON UNIVERSITY GRADUATE SCHOOL OF ARTS AND SCIENCES Thesis REAL-TIME 4D TUMOR TRACKING AND MODELING FROM INTERNAL AND EXTERNAL FIDUCIALS IN FLUOROSCOPY

Fluoroscopy is currently used in treatment planning for patients undergoing radiation therapy for lung and abdominal tumors. Radiation oncologists seek to maximize the radiation the tumor receives, and minimize the amount delivered to the surrounding tissues, a task made difficult by movement caused by the breathing of the patient. A model of the tumor motion could greatly improve dose calculation and delivery. There exists a system which directly estimates the location of the tumor, but this system is very costly and not widely deployed. In this thesis a model which can be derived using less specialized equipment is presented. First, a method of tracking the two-dimensional (2D) motion of internal markers (surgical clips) placed around the tumor is presented and examined. A ground truth is established by visual inspection of 10 data sets of 5 patients to evaluate the tracker. The root mean squared error in estimating 2D marker position was 0.47 mm on average. Using two orthogonal sequentially obtained fluoroscopic image sequences, a method for calculating a model of the average or maximum three-dimensional (3D) motion of the clips is presented, examined, and compared to the direct estimation system. On average, the error was 3.0 mm for four pairs of trajectories. If imaging is possible during treatment, this modeled motion can be used for beam-guided radiation, otherwise, the modeled motion iv can be correlated to a set of external markers for use in respiratory gating.