Spatio-temporal motion prediction model for liver cancer radiotherapy

自由呼吸的肝癌疗法，如外照射疗法(EBRT)，需要在治疗过程中变形和移动的器官内进行准确的肿瘤跟踪。然而，肿瘤靶向干预的主要局限性在于患者的呼吸或非自愿运动，这可能会偏离预定的目标和在计划期间确定的轨迹与实际解剖结构，从而导致执行该行动的治疗设备相对于目标的相对位置上的误差。内部解剖的实时运动跟踪依赖于3D成像和图像的实时后处理，这在介入手术中是不可行的。因此，为了在治疗期间补充临床上可用的部分信息(2D图像、导航信号)，需要预先了解呼吸周期中的预期运动场。该项目的目标是开发和评估一个基于深度学习和生物力学模型的时空肝脏运动模型，该模型将预测肿瘤在呼吸周期内的移位，并帮助放射肿瘤学家不仅跟踪肿瘤，而且避免损害肿瘤靶点周围的关键结构。该项目将产生显著的直接和间接经济效益。至于直接的好处，主要的工业合作伙伴-Elekta是放射治疗和放射外科设备和软件的创新者，将可以访问4D-MRI成像数据和基于机器学习的计算机化模型，这将满足他们在治疗期间对运动管理的关键需求。他们与飞利浦医疗保健公司合作开发的MR-Linac系统将对结果特别有利，以提高靶向准确性，并将辐射过程中的器官变形伪影降至最低。间接好处包括更好的病人护理和减少所需干预的数量。这一多学科的翻译研究计划将为生物医学领域的2名研究生、3名本科生和1名博士后研究员提供与医学成像和病理学专家合作的独特机会，并为医学生、住院医生和研究员提供翻译临床领域基本概念的独特机会。

Free-breathing liver cancer therapies such as external beam radiation therapy (EBRT) require accurate tumor tracking within an organ which deforms and moves during treatment. However, a major limitation of tumor-targeted interventions resides in the patient's respiration or involuntary movement, which may stray the pre-defined target and trajectories determined during planning from the actual anatomy, thus inducing errors in the relative position of the therapy device performing the action with respect to the target. Live motion tracking of the internal anatomy depends on 3D imaging and image post-processing in real-time, which is unfeasible during interventional procedures. Thus, to complete partial information (2D images, navigator signal) clinically available during treatment, prior knowledge of the anticipated motion field during the breathing cycle is necessary. The project objective is to develop and evaluate a spatio-temporal liver motion model based on deep learning and biomechanical models, which will predict tumor displacement within the breathing cycle and help radiation oncologists not only track tumors, but also avoid damaging critical structures surrounding the tumor target. This project will have significant direct and indirect economic benefits. As to the direct benefits, the primary industrial Partner - Elekta - who is an innovator of equipment and software for radiation therapy and radiosurgery, will have access to 4D-MRI imaging data and computerized models based on machine learning, which will address their critical needs for motion management during therapy. The outcome will be particularly beneficial with their MR-Linac system, developed in partnership with Philips Healthcare, in order to improve targeting accuracy and minimize organ deformation artefacts during radiation. Indirect benefits include better patient care and reduced number of required interventions. This multidisciplinary translational research program will provide a unique opportunity for 2 graduate students, 3 undergraduate students and one post-doctoral fellow in the field of biomedical sciences to work with experts in medical imaging and pathology and for medical students, residents and fellows to translate fundamental concepts in the clinical field.**********