Motion Modelling and Motion Compensated Reconstruction from Cone-Beam CT Projection Data

1)研究背景的简要描述，包括潜在影响大多数放射治疗机器上都有机载锥形束CT（CBCT）成像系统，但图像采集时间约为1分钟，难以重建显示呼吸运动的图像。该研究项目将开发直接从未重建投影数据中建模呼吸运动的方法，并使用这些模型对CBCT体积进行运动补偿重建。这可能为肺癌患者带来更准确、更安全和更有效的治疗。2）目的和目标1.开发、实施和评估可应用于CBCT投影数据的“替代驱动”和“无替代”运动模型。研究CBCT数据运动补偿重建的不同方法，包括现代迭代和基于学习的方法.研究基于现代机器学习的方法，将运动模型拟合到投影数据中。4.建立使用运动模型的工具和工作流程，以通知和指导自适应放射治疗。3)研究方法的新奇在UCL和其他地方已经进行了初步的工作，直接将运动模型拟合到投影数据，但到目前为止，这还没有成功地应用于真实的患者数据，还有一些挑战有待克服。此外，所有以前的工作都集中在“代理驱动”的模型。使用低秩分解方法的无替代运动模型最近已被提出用于MRI，但以前尚未应用于CBCT数据。还将研究新的方法，用于执行运动补偿重建和使用现代机器学习方法拟合运动模型。最后，运动模型将使新的方法被开发，以适应放射治疗，以更好地解释呼吸运动。4）与EPSRC的战略和研究领域的一致性该项目与EPSRC的医疗保健技术主题一致，特别是扩展物理干预的前沿和优化疾病预测，诊断和干预的挑战。它也与医学成像研究领域保持一致。5)Elekta将参与该项目。克里斯蒂医院和癌症研究所也将通过提供CBCT数据和专家建议来合作该项目

1) Brief description of the context of the research including potential impactOn-board Cone-Beam CT (CBCT) imaging systems are available on the majority of Radiotherapy treatment machines, but image acquisition times of ~1 minute make it difficult to reconstruct images showing respiratory motion. This research project will develop methods to model the respiratory motion directly from unreconstructed projection data, and use these models to perform a motion-compensated reconstruction of the CBCT volume. This could potentially lead to more accurate, safer, and more effective treatments for lung cancer patients.2) Aims and Objectives1. Develop, implement, and evaluate 'surrogate-driven' and 'surrogate-free' motion models that can be applied to CBCT projection data.2. Investigate different methods for performing the motion-compensated reconstruction of the CBCT data, including modern iterative and learning based approaches.3. Investigate modern machine learning based approaches for fitting the motion models to the projection data.4. Establish tools and workflows for using the motion models to inform and guide adaptive radiotherapy. 3) Novelty of Research MethodologyInitial work has been undertaken at UCL and elsewhere on fitting motion models directly to projection data, but so far this has not been successfully applied to real patient data and there are a number of challenges still to be overcome. Furthermore, all previous work has focussed on 'surrogate-driven' models. Surrogate-free motion models using a low-rank decomposition approach have recently been proposed for MRI but have not previously been applied to CBCT data. Novel methods will also be investigated for performing the motion compensated reconstructions and for fitting the motion models using modern machine learning approaches. Finally, the motion models will enable novel methods to be developed for adapting radiotherapy treatment to better account for the respiratory motion.4) Alignment to EPSRC's strategies and research areasThis project is aligned with EPSRC's Healthcare technology theme, especially the challenges of Expanding the Frontiers of Physical Intervention and Optimising Disease Prediction, Diagnosis, and Intervention. It is also aligned with the Medical Imaging research area. 5) Any companies or collaborators involvedElekta will be involved in the project.The Christie Hospital and the Institute of Cancer Research will also collaborate on the project by providing CBCT data and expert advice