Development of a tool to extract quantitative image features and predict outcome

描述（由申请人提供）：项目概述众所周知，临床参数（如临床分期）与癌症患者的生存结局相关。然而，患者之间存在很大的差异，我们无法准确或可靠地预测个体患者的生存期-这是个性化癌症医学的必要步骤。对于患者、护理人员和临床工作人员来说，非常需要准确、可靠的结果预测。即使是基因数据的加入，也没有在临床上显著增加我们对个体患者结局预测的可靠性。最近的研究，包括我们自己的研究，已经表明，从治疗前CT图像中提取的图像特征可用于预测非小细胞肺癌患者，食管癌患者等的治疗结果。目前研究的一个局限性是缺乏一个通用平台，使研究能够共享结果，并快速轻松地将技术应用于自己的患者数据集。我们提出的项目将创建开源软件工具，这些工具将与当前可用于放射治疗研究的开源工具集成。我们还将深入研究计算图像特征所涉及的各种不确定性来源，使研究人员能够避免使用对成像参数（如像素大小）具有高度依赖性的特征。近100%的NCI资助的临床试验包括治疗前CT成像。我们的初步工作将提供工具，让参与这些研究的研究人员调查使用定量图像特征预测治疗结果。

DESCRIPTION (provided by applicant): Project Summary It is well known that clinical parameters such as clinical stage are correlated with survival outcomes among cancer patients. However, there is much variability among patients, and we are unable to accurately or reliably predict survival for individual patients - a necessary step for personalized cancer medicine. There is a strong need for accurate, reliable outcome predictions for patients, caregivers, and clinical staff. Even the addition of genetic data has yet to make a clinically significant increasein the reliability of our outcome prediction for individual patients. Recent research, including our own, has shown that image features extracted from pre-treatment CT images can be used to predict treatment outcomes for non-small cell lung cancer patients, esophageal cancer patients, and others. A limitation to current studies is the lack of a common platform that would enable research to share results and quickly and easily apply techniques to their own patient datasets. Our proposed project will create open-source software tools that will integrate with current open-source tools that are available for radiation therapy research. We will also carry out an in-depth investigation into the various sources of uncertainty involved in calculating image features, allowing researchers to avoid using features that have high dependence on imaging parameters (such as pixel size). Nearly 100% of NCI- funded clinical trials include pre-treatment CT imaging. Our preliminary work will provide the tools to allow researchers involved in these studies to investigate the use of quantitative image features for predicting treatment outcome.