Development, Validation and Applications of MRI-Only Treatment Planning in Radiotherapy

Abstract

Magnetic resonance imaging (MRI) has superior soft tissue visualization to guide radiotherapy treatment planning but does not provide the electron density information required for the dose calculation. Thus, MRI has been used in a complementary way, registering to the gold standard computed tomography (CT) scan. Development of methods to allow accurate planning from the MRI images would remove the requirement for additional (CT) scans as well as improve clinical workflow and remove potential registration errors. Various methods have been reported to generate datasets with electron density information from MRI data, with these being termed substitute, synthetic or pseudo CT (sCT) datasets.

This thesis explores the potential variation in planning and optimization error from MRI-only treatment planning for a range of situations. sCT generation was explored with a deep learning methodology applied to a set of retrospective H&N patient data. A lung MRI sequence was investigated for its potential application for sCT generation, with various methods trialed and assessed for clinical suitability. For an existing sCT generation method used clinically for prostate cancer treatment planning, a time-reduced MRI sequence was investigated, optimizing scan parameters for this by initial assessment in a volunteer cohort, followed by clinical validation in a patient cohort. A pancreas MRI volunteer study was also conducted to investigate internal organ motion effects on treatment planning and potential treatment delivery to assess the suitability of treatment regimes for pancreatic cancer patients.

This work provides evidence that MRI-only treatment planning is achievable and acceptably accurate. This has led to current and future implementations of findings into clinical practice locally, and potentially more widely. MRI-only treatment planning in radiotherapy could lead to improved patient outcomes, via both better target delineation and reduced normal tissue toxicity.