面向肿瘤治疗监测的多中心PET/CT影像的定量分析研究

Because of its potential in prediction of cancer treatment outcome, PET/CT imaging performed in cancer treatment monitoring will be helpful for personalizing cancer treatment strategies. Considering the benefit of PET/CT imaging in cancer therapy, this project begins with benefit evaluation based on multicenter clinical data, and then studies the quantitative analysis method and the properties of quantitative parameters in order to enhance the value of PET/CT in cancer treatment. This project consists of three steps: (1) To quantify the impact of PET/CT during cancer treatment: statistical analysis of clinical questionnaires is applied for evaluating the benefit of PET/CT in treatment monitoring. (2) To establish standard quantitative analysis method for multicenter data: this method integrates with automatic region of interest delineation using statistical model-based image segmentation algorithms, modification of standard uptake value based on the properties of in-vivo tracer distribution, and non-invasive quantitative analysis method with image-derived input function. (3) To explore optimal feature of treatment response and its relation model with patient's outcome: the quantitative parameters from PET/CT images, taken before and after therapy, will be compared. The parameter or combination of parameters, which is sensitive to tumor response will be extracted by data mining and referred as feature of treatment response. The relation between the feature and patient's outcome will be modeled to provide a new basis for evaluating therapy and predicting treatment outcome. This will further improve the development of personalized cancer treatment with assistance of PET/CT.

PET/CT显像具有早期预测肿瘤治疗效果的潜能，用于肿瘤治疗监测将有助于肿瘤治疗的个性化。围绕PET/CT显像的这一效益，本项目基于多中心临床数据从效益论证着手，开展定量分析研究并探索定量参数特性，以提升PET/CT显像在肿瘤治疗中的价值。该研究分为三步：①量化PET/CT在肿瘤治疗中的影响力，通过临床问卷统计分析评估PET/CT显像用于肿瘤治疗监测的效益；②建立多中心影像的定量分析方法。其中包括：采用基于统计模型的图像分割算法实现感兴趣区自动定位，根据显像剂分布特性改进标准摄取值计算，及应用影像获取输入函数进行无创定量分析；③探索优化的疗效特征及其与预后的关系模型。比较治疗前后PET/CT影像定量参数，通过数据挖掘提取对肿瘤反应敏感的定量参数作为疗效特征，并建立疗效特征与患者预后的关系模型，以期为早期疗效评估和预后预测提供新依据，并促进PET/CT辅助个性化肿瘤治疗策略制定的发展。

本项目根据研究计划，我们先后对核医学影像的定量分析的方法理论及实际应用进行了探索。为了便于多中心多模态影像数据的收集、存储与分析，在本项目研究中搭建了基于互联网技术的数据收集平台，并在项目执行过程中加以完善。针对核医学影像高噪声低分辨率的特点，结合多模态影像的优势，从改进图像分割算法入手，研究用于PET/CT及SPECT/CT等功能影像的快速稳定且客观的感兴趣定位方法。该项研究结果不仅有助于病灶区域定位，同时也有助于指导寻找治疗后恢复区域，对于疾病的治疗监测有着重要的作用。充分利用核医学影像的功能信息，展开多种功能影像定量分析方法研究。针对SPECT影像，引入归一化方法实现不同患者不同时间数据的比较，为治疗监测提供了一种便捷定量的方法分析时序影像，所获得的定量参数有利于疗效评估；针对PET动态图像参数估计的需求，对短时动态PET显像策略及定量参数估计进行了研究，有助于提高临床或临床前PET动态显像的效率和通量；发现利用动态FDG-PET图像动力学参数可以较好地区分恶性肿瘤和炎症组织，研究结果有助于恶性肿瘤的诊断与评级。在此基础上，借助核医学影像定量分析方法进行治疗监测疗效评估研究，对多种疾病的治疗方案选择及治疗监测进行了探索，有助于提高临床使用核医学影像进行疾病诊治和疗效评估的及时性、客观性和可靠性。