Towards personalized medicine with theranostics: quantitative molecular imaging and artificial intelligence

核医学中令人兴奋的治疗声学(即治疗+诊断)领域涉及将放射性药物成像与放射性药物治疗(RPT)结合使用(之前、期间或之后)。成像是用单光子计算机发射断层扫描/计算机断层扫描(SPECT/CT)或正电子发射断层扫描(PET)进行的，而RPT使用发射α或β辐射的放射性同位素。与常规的体外放射治疗(EBRT)相反，RPT目前没有根据患者的个体特征进行计划。RPT中的剂量测定被认为是困难和耗时的，因此，传递到肿瘤/健康器官的辐射剂量通常没有量化。涉及物理学、应用数学、工程学和计算机科学的基础科学研究，在与医生和技术专家密切合作的同时，对于缩小RPT和EBRT之间的差距至关重要。我提出的计划旨在利用定量成像和人工智能(AI)将RPT与EBRT相提并论，以允许RPT中的个性化剂量评估。我的短期目标是：1)开发正电子发射计算机断层扫描和SPECT成像方案，以测量放射性药物的生物分布，即使是对不能直接成像的放射性同位素也是如此。2)开发稳健的图像分割算法，从SPECT/CT和PET/CT图像中准确地量化肿瘤/器官放射性和质量(剂量学中都需要)，并使用人工智能实现自动分割。3)研究定量图像生成和剂量测量方法(例如器官水平与体素方法)对剂量估计的不确定度的影响。4)将诊断PET、治疗SPECT和剂量图像的定量、纹理、形状和强度图像特征(放射组学)结合起来，以改进/简化剂量估计并建立剂量/反应的预测模型。我的计划在培养高素质人才(HQP)方面具有很强的针对性。HQP被指导成为专家/领导者，无论他们是在工业、学术界还是医疗保健实践中追求职业生涯。HQP将在积极指导和支持的同时发展自己的思维和决策。他们参与并加强了我们的产学研合作。HQP将在科学会议上建立网络并传播我们的工作，并鼓励他们加入我们领域的协会委员会。将在一个多学科、多文化、多样化和包容性的团队环境中追求和推动科学的卓越。我们在定量成像、剂量测定、人工智能方面的经验，以及不列颠哥伦比亚省癌症大学和不列颠哥伦比亚大学的支持，使该项目成为将加拿大推向个性化RPT前沿的绝佳场所和机会。该计划将为简化剂量测定协议、建立自动分割工具、实现个性化RPT规划、培训HQP以及扩大与行业和学术合作伙伴的合作做出重要贡献。

The exciting field of theranostics (i.e. therapy + diagnostics) in nuclear medicine involves usage of radiopharmaceutical imaging in tandem with (before, during or after) radiopharmaceutical therapy (RPT). Imaging is done with single photon computed emission tomography/computed tomography (SPECT/CT) or positron emission tomography (PET), while RPT uses alpha- or beta-emitting radioisotopes. Contrary to what is routine for external beam radiation therapy (EBRT), RPT is presently not planned based on individual patient characteristics. Dosimetry in RPT is believed to be difficult and time-consuming, and thus, radiation doses delivered to tumors/healthy organs are commonly not quantified. Basic science research involving physics, applied math, engineering, and computer science, while working closely with physicians and technologists, are essential in closing the gap between RPT and EBRT. My proposed program aims at bringing RPT on par with EBRT, utilizing quantitative imaging and artificial intelligence (AI), to allow for personalized dose assessments in RPT. My short-term objectives are: 1)Develop PET and SPECT imaging protocols to measure radiopharmaceutical biodistribution using theranostic pairs even for radioisotopes that cannot be directly imaged. 2)Develop robust image segmentation algorithms to accurately quantify tumor/organ radioactivity and mass (both required in dosimetry) from SPECT/CT and PET/CT images, and enable automated segmentation using AI. 3)Study contributions to uncertainties in the dose estimates from quantitative image generation and dosimetry method (e.g. organ level vs. voxelized methods). 4)Combine quantitative, texture, shape, and intensity image features (radiomics) from diagnostic PET, therapy SPECT, and dose images to improve/simplify dose estimates and build predictive models for doses/responses. My program is designed with a strong component in training of highly qualified personnel (HQP). HQP are mentored to become experts/leaders, whether they pursue a career in industry, academia, or healthcare practice. HQP will develop their own thinking and decision-making, while actively mentored and supported. They participate in and enhance our industry-academic collaborations. HQP will network and disseminate our work at scientific conferences and are encouraged to join committees of societies in our field. Scientific excellence will be pursued and propelled in the context of a multidisciplinary, multicultural, diverse, and inclusive team environment. Our experience with quantitative imaging, dosimetry, AI, and the support from both BC Cancer and the University of British Columbia make this program a great venue and opportunity to put Canada at the forefront of personalized RPT. The program will make critical contributions to enable simplified dosimetry protocols, establish tools for automatic segmentation, enable personalized RPT planning, train HQP, and expand collaborations with industry and academic partners.