Navigating MRI-guided lung radiotherapy into the fourth dimension

In 2022, lung cancer had both the highest incidence and the highest mortality in both sexes combined. Image-guided and adaptive radiotherapy has emerged as a treatment option for inoperable lung cancer patients. One of the major uncertainties in radiotherapy of lung tumors is the breathing-induced motion. The current standard of care to account for breathing motion is respiratory-correlated four-dimensional computed tomography (4D-CT), which represents the three-dimensional (3D) anatomy in typically ten different breathing phases. Although magnetic resonance imaging (MRI) is increasingly used as an imaging modality for radiotherapy, no 4D-MRI methodology yet exists for MR-guided radiotherapy at MR-linac systems. The aim of this thesis was to investigate the development of a 4D-MRI methodology on the 1.5 T Unity MR-linac. This development would enable motion management on the day of treatment, which goes beyond the adaptive radiotherapy workflows currently available that are limited to correcting for daily anatomical changes based on a 3D-MRI scan.

The basis of this thesis was the development of an MRI methodology consisting of the acquisition sequence and the methodology to process the images into a respiratory-correlated 4D-MRI. A standard two-dimensional (2D)-MRI sequence was modified to acquire two slices simultaneously to extend the coverage of the anatomy in the slice direction. After processing the data into a 4D-MRI, they were subsequently transformed into a time-weighted average 3D-MRI (i.e., mid-position) that is compatible with the treatment planning software.

The trade-off between the spatial resolution and acquisition time resulted in a limited accuracy for treatment planning when using this mid-position. Therefore, the spatial resolution of the mid-position MRI was improved by deforming a high-resolution MRI scan to the mid-position based on 4D-MRI-derived motion information. In addition to the use for treatment planning and adaptation, it was also successfully demonstrated that the 4D-MRI methodology could be used for real-time motion estimation and target tracking during treatment (the latter in a phantom).

To improve the representativeness of the 4D-MRI for the underlying average breathing and to minimize the risk of target drift, a visual biofeedback method using an internal navigator was developed. Breathing regularity and stability were drastically improved in healthy volunteers. As a result, a shorter scan time could suffice for the required amount of data to correctly represent the average breathing cycle with the 4D-MRI. In the last chapter, an improved visual biofeedback implementation was used to guide healthy volunteers and one patient during the acquisition of different 4D-MRI sequences in an international collaboration, allowing for in vivo comparisons between different 4D-MRI methods.

In conclusion, the research presented in this thesis introduced a 4D-MRI methodology that enables motion estimation for treatment planning, can define the time-weighted tumor position for treatment planning, and can be used for real-time motion estimation during irradiation. When combined with visual biofeedback, breathing is regularized and stabilized, which improves the representativeness of the 4D-MRI and derived time-weighted average anatomy. All-in-all, enabling 4D-MRI as motion management at the MR-linac will further expand the patient-specific treatments in MR-guided radiotherapy and hopefully lead to improved patient outcomes.