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

肺癌是英国第三大常见癌症，每年约有45，000名新患者。生存率仍然很低，三分之一的患者经历局部失败，5年生存率仅为9.5%。放射治疗是治疗的重要组成部分，给予>50%的患者。研究表明，提高肿瘤剂量可以改善局部控制，但是对辐射敏感的附近正常组织，如心脏，限制了可以安全输送的放射治疗剂量。大多数患者每天接受20-33次放疗。在每次分割之前，必须对患者进行成像，以确保其相对于预期治疗光束正确定位。这也允许在必要时调整治疗，例如，如果患者体重减轻。最近推出的磁共振直线加速器（MR-Linac）提供同时的MR成像和辐射。这使患者能够受益于MRI相比CT的上级图像质量，从而提高了疾病的靶向和健康组织的保留。它还提供高质量的实时成像，以在照射过程中跟踪肿瘤。MR-Linac的主要限制是治疗时间长，由于目前需要复杂的工作流程，每次治疗时间可能超过一小时。尽管MRI设置图像为治疗和保留重要器官提供了更好的可视性，但在标准直线加速器上获取这些图像所需的时间比CBCT长。在治疗床上花费长时间对于患者来说是不舒服的，并且增加了患者移动的风险，这可能使得位置校正花费更长时间。因此，临床上迫切需要实现新的成像技术，其可以加速机载图像采集，但没有通常由较短的成像时间和运动引起的伪影。该项目将把最新的加速磁共振成像概念应用于肺癌，并将这些技术应用于临床实践。在这个项目中，学生将：1）使用JEMRIS模拟器创建一个数字胸部模型，可用于测试不同的成像技术。2）开发一种基于深度学习的方法，用于回顾性脂肪抑制。3)开发基于MR的视觉引导屏气系统。4)评估视觉引导对屏气部分内重复性的影响。

Lung cancer is the third most common cancer in the UK with around 45,000 new patients per year. Survival remains poor, with a third of patients experiencing local failure and a 5 year survival of only 9.5%. Radiotherapy is an important component of treatment, given to >50% of patients. Studies have shown that elevating tumour dose can improve local control, however nearby normal tissues that are sensitive to radiation, such as the heart, limit the radiotherapy dose that can be delivered safely. Most patients receive radiotherapy in 20-33 daily fractions. Before each fraction the patient must be imaged to ensure they are correctly positioned relative to the intended treatment beams. This also allows the treatment to be adjusted if necessary, for example if the patient has lost weight.The recently introduced magnetic resonance linear accelerator (MR-Linac) provides simultaneous MR imaging and radiation. This allows patients to benefit from the superior image quality of MRI compared to CT, which improves the targeting of disease and sparing of healthy tissue. It also offers high quality, real-time imaging to track the tumour during irradiation.The primary limitation of the MR-Linac is the long treatment times, which can exceed an hour per fraction, due to the complex workflow currently required. Although MRI setup images offer improved visibility for treating and sparing of essential organs, acquiring these images takes longer than CBCT on a standard linear accelerator. Spending a long time on the treatment couch is uncomfortable for the patient and increases the risk of movement of the patient which can make the position correction take even longer. There is therefore an urgent clinical need for the implementation of novel imaging techniques which can accelerate on board image acquisitions, but without the artefacts normally caused by shorter imaging times and motion. This project will apply the latest concepts in accelerated MR imaging to lung cancer and bring these techniques into clinical practice. In this project, the student will:1) Use the JEMRIS simulator to create a digital thorax model which can be used to test out different imaging techniques.2) Develop a deep learning based method for retrospective fat suppression. 3) Develop an MR-based system for visually guided breath holds. 4) Assess the impact of visual guidance on intra-fraction repeatability of breath holds.