Radiotherapy aims to deliver a high radiation dose to the target, while sparing the healthy surrounding tissues. However, treatment fails for almost half of head and neck cancer patients. Therefore, attention is directed towards using improved imaging techniques such as PET and MRI for radiotherapy planning.
The aim was to enable the use of the multiparametric PET/MRI scanner as a one-stop shop for radiotherapy planning in head and neck cancer patients.
We designed a radiotherapy setup for PET/MRI and demonstrated the clinical feasibility of performing multiparametric PET/MRI of patients in radiotherapy treatment position.
We developed a deep learning network for deriving synthetic CT from MRI and evaluated it for radiotherapy dose calculations and PET attenuation correction, while keeping the real planning CT as the reference. The results showed only small errors - even when the model was applied to completely independent data from another institution.
We further demonstrated the importace of performing distortion correction of diffusion weighted MRI, if multiparametric imaging is to be used for more individualized radiotherapy planning.
The studies presented in this thesis (and linked below) have contributed to the enabling of multiparametric PET/MRI for radiotherapy planning and may open up for new treatment possibilities.
Place of employment
Dept. of Clinical Physiology, Nuclear Medicine and PET
Date and place of defense
23rd June 2021, Auditorium 1 (Entrance 44), Rigshospitalet
Prof. Andreas Kjær
Prof. Barbara Malene Fischer
Assoc. Prof. Flemming Littrup Andersen
Prof. Adam Espe Hansen
Dr. Jacob Høygaard Rasmussen