Next-generation image analytics for chemotherapy and survival prediction in pancreatic ductal adenocarcinoma

该项目的目标是建立一个强大的、质量控制的图像处理管道，并从标准化的常规临床计算机断层扫描成像中生成、验证和测试PDAC患者化疗反应和生存的预测模型。研究目标我们的具体目标是：(1)建立一个图像预处理和分割流水线，包括一个基于种子的区域增长算法，能够自动分割胰腺癌病变，可以部署到客户端设备，供经验有限的胰腺癌放射科医生使用；(2)基于基线和早期随访影像数据，建立基于放射组学的模型，预测当前强化化疗方案的短期和长期反应；(3)创建一个多变量无病/总生存模型，包括衍生的放射组学特征和NEOLAP试验的临床数据，以评估放射组学方法相对于传统风险评估工具的增量价值。我们期望我们的研究：(1)通过识别与治疗相关的肿瘤亚型的不同放射学特征来提高我们对PDAC生物学的理解；(2)确定预测PDAC特异性治疗反应和个体生存的放射学特征；(3)为非专家放射科医生和临床合作伙伴提供有关患者分层（即个性化医疗）和治疗监测的决策支持工具；(4)促进胰腺癌和精准肿瘤学领域的研究，为预测患者预后提供先进的图像分析算法。

The objective of the project is to establish a robust, quality-controlled image processing pipeline and to generate, validate and test predictive models of chemotherapy response and survival in patients with PDAC from standardized routine clinical computed tomography imaging.Research AimsFollowing the objectives, we aim specifically: (1) to establish an image pre-processing and segmentation pipeline including a seed-based region growing algorithm capable of automatically segmenting pancreatic cancer lesions which can be deployed to client-side devices for usage by radiologists with limited experience in pancreatic cancer;(2) to generate a radiomics-based model predictive of short- and long-term response to current intensified chemotherapy regimens based on baseline and early follow-up imaging data; (3) to create a multivariate disease-free/ overall survival model including the derived radiomic signatures alongside clinical data from the NEOLAP trial to assess the incremental value of a radiomics approach over traditional risk assessment tools.Expected ImpactWe expect our research:(1) to improve our understanding of PDAC biology by identifying distinct radiomic features of treatment-relevant tumor subtypes;(2) to identify radiomic signatures predictive of specific therapy response and individual survival in PDAC;(3) to provide non-expert radiologist and clinical partners with tools for decision support regarding patient stratification (i.e. personalized medicine) and therapy monitoring;(4) to facilitate research in the field of pancreatic cancer and precision oncology, providing advanced image-analysis algorithms for the prediction of patient outcome.