Methods for fast and accurate MR-guided radiotherapy in lung cancer

Abstract

Lung cancer is one of the most common types of cancer, but treatment outcomes remain poor. Magnetic Resonance Guided Radiotherapy (MRgRT) offers better soft tissue contrast, allowing more accurate targeting of the tumour and sparing of healthy tissues. However further work is required to optimise MRgRT for the treatment of lung cancer. In this thesis a toolbox of methods for reducing imaging time and dealing with respiratory motion are described. Lung anatomy and respiratory motion present challenges for lung MRI and require development of bespoke imaging sequences for MRgRT. Bloch simulation of MR acquisition is a useful tool for testing prototype sequences, but no available software provides anatomical structure, realistic breathing patterns and deformable motion in one package. A 4D digital phantom framework was developed to provide these features. Its use was demonstrated through simulating 2D cine of the lungs and expected motion ghosts in a T2-weighted spin echo acquisition. MRI allows acquisition of images with different contrast without the use of intravenous contrast agents. As acquiring multiple images is time-consuming, a Cycle-Consistent Adversarial Network (cycleGAN)-based retrospective fat suppression method was developed. It successfully reduced the intensity of the fat signal, without substantially changing the tumour signal intensity. This method reduces imaging time by 6 minutes when images with and without fat suppression are required. Breath holds are used as a motion management technique in imaging and treatment. However patients can struggle with performing serial breath holds in a repeatable and reproducible manner. An MRI-navigator-based visually guided breath hold method was developed and tested on three patient volunteers. Visual guidance improved the reproducibility and stability of breath holds in two out of three patients, reducing inter-breathhold motion up to 50%. Together these methods serve as a toolkit for researchers working towards making MRgRT of lung cancer more efficient and accurate.

Date of Award	18 Jul 2024
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Benjamin Rowland (Supervisor), Cynthia Eccles (Supervisor) & Marcel Van Herk (Supervisor)