Author ORCID Identifier
Date of Graduation
Doctor of Philosophy (PhD)
Clifton David Fuller, MD, PhD
Kristy Brock, PhD
Geoffrey Ibbott, PhD
Mary Martel, PhD
Jihong Wang, PhD
Ying Yuan, PhD
The 1.5T hybrid MRI/linear accelerator (MR-linac) has recently been introduced into clinical practice and used for the treatment of head and neck cancers (HNC). This device enables on-line adaptive radiation therapy (ART) based on anatomical changes throughout treatment and variations in patient position. This novel technology also has the potential for advanced ART strategies such as dose-optimized ART, in which the treatment plan is optimized based on the accumulated dose over previous fractions, or biological image-guided ART, in which the plan is adapted based on individual tumor response as measured through quantitative imaging techniques such as diffusion-weighted imaging (DWI). The aims of this dissertation are to validate the existing adaptive workflows for HNC on the MR-linac and perform technological validation of preliminary steps for these advanced ART strategies.
First, we measured the dose distribution caused by the electron return effect at the interface of high- and low-density materials using gel dosimetry. The second project focused on validating our clinical workflow for the first ten HNC patients treated on the MR-linac. We demonstrated that we could create quality adaptive treatment plans for HNC but that the poor autosegmentation performance is a major bottleneck of the on-line Adapt to Shape workflow.
Next, with the ultimate goal of creating a dose accumulation tool for MR-linac treatments, we developed a method for reconstructing the delivered dose from Adapt to Position plans. Because these doses are calculated on the reference image rather than the setup image to save time in the on-line workflow, doses can be recalculated on the setup image off-line after tumor and OAR segmentation to determine the true delivered dose. We evaluated the performance of various autosegmentation algorithms on the MR-linac setup images and investigated how the segmentation accuracy impacts the dose calculation.
The last component of this project was optimization and technical validation of DWI sequences on the MR-linac. We optimized turbo spin echo (TSE) and split acquisition of fast spin echo signals (SPLICE) using both quantitative and qualitative metrics. Finally, we compared these sequences to echo planar imaging (EPI) on the MR-linac and three DWI sequences on a 1.5T MR simulator by measuring in vivo repeatability, ADC bias, and signal-to-noise ratio.
In conclusion, these projects validate the clinical feasibility of treating HNC on the MR-linac and pave the way for advanced adaptive strategies such as dose accumulation and biological image-guided ART to personalize radiation therapy for HNC patients.
MR-guided adaptive radiation therapy, diffusion-weighted imaging, head and neck cancer, MR-linac