MR-only and PET-MR Radiotherapy for Pelvic Cancers

Abstract

42,500 people are treated each year for cancers in the pelvis in the UK. Radiotherapy is a key treatment technique for pelvic cancers, with approximately 40% of patients receiving it as the primary or adjuvant treatment. Critical to targeting the radiation accurately is the quality of the imaging used to plan radiotherapy treatments. Conventional Computed Tomography (CT) imaging is geometrically robust and provides mass density information for accurate radiotherapy dose calculations. However in the pelvis the poor soft-tissue contrast of CT makes delineation of the tumour and nearby healthy Organs At Risk (OARs) difficult and CT is unable to provide functional or metabolic information about the tumour. In contrast, Magnetic Resonance (MR) has superb soft-tissue contrast, improving the accuracy of tumour and OAR delineation, and is also able to provide additional functional information to further characterise the tumour such as Diffusion Weighted (DW)- MR. When combined with the metabolic information available from Positron Emission Tomography (PET) in a simultaneous PET-MR scanner, this has great potential to enable the identification of tumour sub-volumes for receiving boost radiation doses and to characterise the tumour for more stratified radiotherapy dose prescriptions. However there are significant scientific and technical barriers to using MR-only and PETMR imaging for radiotherapy planning in the pelvis. The aim of this thesis was to develop technical solutions to enable MR-only and PET-MR for radiotherapy planning of pelvic cancers and to evaluate these solutions for clinical radiotherapy treatments. The primary barrier to MR-only radiotherapy is that MR images cannot be used directly for radiotherapy dose calculations. This dissertation describes the development of a synthetic CT (sCT) Deep Learning model based on a novel zero echo time MR sequence, in collaboration with GE Healthcare, and its comprehensive evaluation for a range of pelvic radiotherapy treatments. Additionally, a separate Deep Learning algorithm that automatically contoured OARs, also developed by GE Healthcare, was evaluated for prostate, anal and rectal cancer sites. Finally, clinical implementation of MR-only radiotherapy also requires a method for ongoing Quality Assurance (QA) of the sCT dose calculation accuracy. A method using Cone Beam (CB)CT was developed and analysed on a cohort of clinical MR-only patients. A major barrier for the use of PET-MR imaging for pelvic radiotherapy is the impact on both PET and MR image quality when acquiring images in the pelvic radiotherapy position. This image quality loss was quantified using phantoms and methods of incorporating the radiotherapy hardware into the PET Attenuation Correction (AC) map were developed. The impact of using these AC maps on tumour delineation and metabolic characterisation was then investigated in anal and rectal radiotherapy patients. Acquiring an accurate PET image also requires an AC map of the patient. This is challenging for PET-MR because MR, unlike CT, cannot directly be used for PET AC. This i could be overcome by using the MR-generated sCT evaluated for MR-only radiotherapy to generate a patient AC map. The accuracy of PET images reconstructed using sCTAC compared to gold standard CTAC and also the current MR-based MRAC was evaluated. Another essential component to enabling the use of PET-MR imaging for radiotherapy planning is a QA programme focused on radiotherapy requirements. The tests needed for such a programme were developed and their same-day repeatability and monthly stability over 12 months evaluated. Finally, fully utilising MR and PET-MR imaging for radiotherapy requires imaging times of 20 minutes or more so that high quality anatomical, functional and metabolic information can be acquired. However, for radiotherapy treatment planning it is critical that the positions of the internal anatomy need to be the same during imaging as they are for treatment. During image acquisitions of ≥ 20 minutes, organ motion from changes in bladder filling with be substantial. This organ motion was assessed in healthy volunteers and a method of correcting for the organ motion and developed and evaluated. In summary, this thesis has developed and evaluated methods that enable MR-only and PET-MR imaging to be used for pelvic radiotherapy. These have included automated OAR delineation, bladder filling management, quantitative PET imaging in the radiotherapy position and accurate radiotherapy dose calculation, all underpinned by patient specific and system QA tests. Together, this thesis enables MR-only and PET-MR to be implemented for patients, paving the way for evaluation of their clinical benefits.