Abstract:

Human abdominal organs are subject to a variety of physiological forces that superimpose their effects to influence local motion and configuration. Motions include breathing, gastric contraction, and other types of less periodic slow configuration changes. Breathing motion has been extensively studied and well characterized; however, gastric contraction and slow configuration motion have been rarely investigated. By using a golden angle stack-of-star radial sampling magnetic resonance image (MRI) sequence, we constructed a hierarchical motion model that characterizes each of these three motions, as well as their combined effects. Breathing motion is extracted and corrected as the first step, following by reconstruction of gastric motion and slow configuration changes. The model shows non-neglectable geometric displacements raised by all three motion modes. These motions, if not managed properly during radiation therapy, may potentially result in overdose to normal tissue or underdosage to the tumor target. Magnetic resonance guided radiotherapy (MRgRT) systems have been developed which have the technical capability to address these complex motions, but to date their primary applications have been relegated to management of breathing motion. In this dissertation, we proposed a gastric motion prediction framework to allow real-time management of contractile motion during MRgRT taking advantage of the intra-scan stability of gastric contraction motion observed in patients under standard pre-session eating restrictions. The framework was able to achieve submillimeter prediction error with a sufficient future prediction time to overcome the latency introduced by the image sampling reconstruction, motion assessment and treatment interruption or modification on MR-guided linear accelerators. Motions and deformations during radiation treatment present a challenge to precisely and accurately measure the radiation dose delivered to abdominal organs. A dose accumulation tool, developed based on the hierarchical motion model, was built to estimate dose distributions with abdominal motions. The tool demonstrates potential deviations of dose due to motion and shows exceeding of dose constraints in certain cases. It could support offline adaptation or help record delivered dose more accurately than stationary images used for daily patient positioning and/or online adaptation of treatment plans. The motion model is also currently supporting other clinical applications, including providing improved image quality reconstructions from free-breathing scans for improvement of accuracy of perfusion as well as liver functional maps. In the future, the model can be further utilized in other fields including radiology or gastroenterology.  

Room: LBME 2185 / Zoom Link: https://umich.zoom.us/j/93620134849 Meeting ID: 936 2013 4849 Passcode: 268890

Committee Chair(s): Dr. James Balter and Dr. Rojano Kashani