Abstract:
In recent years, quantitative imaging biomarkers (QIBs) have received an increasing amount of attention in many medical fields in order to improve patient outcome by more personalized treatments. Especially in radiotherapy where computed tomography (CT) and magnetic resonance imaging (MRI) are routinely acquired for treatment planning, great potential of QIBs is evident, but additional patient imaging is time- and cost-intensive over the course of radiotherapy.
Therefore, the newly introduced hybrid MR-Linac system, combining a linear accelerator and an MRI scanner, may provide the possibility of not only more personalized radiotherapy by daily anatomical imaging as well as treatment plan and patient positioning adaptations due to high soft tissue contrast imaging, but also the opportunity of acquiring functional MRI and thereof derived QIBs without any additional logistics. However, the technical capabilities of the 1.5 T MR-Linac are inferior compared to a standard diagnostic MRI scanner, questioning the measurement uncertainties of diffusion-weighted (DW)-MRI and thereof derived apparent diffusion coefficient (ADC) values on the MR-Linac as well as the feasibility of longitudinal DW-MRI acquisitions within MR-guided radiotherapy (MRgRT).
Therefore, the overall aim of this thesis was the technical validation of DW-MRI on the 1.5 T MR-Linac in head and neck cancer (HNC) patients, determining the inherent measurement uncertainties and deviations to a diagnostic scanner in form of repeatability and reproducibility for ADC. Additionally, feasibility of DW-MRI acquisitions sequentially during the course of radiotherapy was investigated while also the changes of mean ADC values as well as ADC-based high-risk subvolumes (HRSs) inside the tumor, potentially describing intratumoral regions with higher radioresistance, were evaluated.
In a first step, the repeatability of mean ADC values derived from DW-MRI on a hybrid MR-Linac was determined in a Test-Retest study. By evaluating repeated DW-MRI acquisitions on the 1.5 T MR-Linac from eleven HNC patients, repeatability of mean ADC values in target volumes as well as normal glandular tissues was determined. Relative repeatability coefficients (relRC) resulted in about 23-31% for primary tumors and lymph nodes, while glandular tissues presented better repeatability of around 15-20%, describing threshold values for discriminating true change from measurement uncertainty. A volume-dependent deterioration of repeatability was also identified in addition.
In a second study including 15 HNC patients treated on the 1.5 T MR-Linac, DW images were acquired before the start of radiotherapy and in week two of treatment on the MR-Linac and a 3 T diagnostic MRI scanner. Reproducibility was assessed comparing the corresponding images from both scanners. We determined within-subject coefficients of variation (wCV) of 13% and 24% for target volumes and glandular tissues. Furthermore, a significant underestimation of mean ADC values from the MR-Linac in comparison to the diagnostic scanner as the gold standard was found. ADC bias was investigated in a phantom experiment where the single-shot spin-echo echo-planar-imaging sequence showed negligible bias from the true phantom values.
Lastly, we evaluated mean ADC values in a patient cohort of 28 HNC patients treated and sequentially imaged on the 1.5 T MR-Linac while also a HRS, defined by a cluster of low ADC values, was analyzed. Here, mean ADC values increased about 50% during the course of radiotherapy, while tumor volumes and HRSs decreased by 45% and 92%, respectively. Hence, sequential ADC measurement and thereof derived HRS analysis is feasible on the 1.5 T MR-Linac which provides the basis for future prospective clinical studies with ADC as a QIB or ADC-based HRS as a radioresistance surrogate for response-adaptive MRgRT.
With this work, the technical framework for the acquisition of DW-MRI and thereof derived ADC values on the hybrid 1.5 T MR-Linac was established as well as the feasibility of ADC values and intratumoral HRS acquisitions sequentially during the course of MRgRT. Therefore, future prospective studies may evaluate the predictive value of DW-MRI for more personalized radiotherapy by treatment adaptations before the start of radiotherapy or through response-adaptive MRgRT, combining the technical framework from this work with the shown prognostic value of DW-MRI.
MR-guided radiotherapy