Quantitative MRI models of HN Cancers for Physiological Adaption of RT

描述(申请人提供)：尽管代谢和生理成像已显示出对放射治疗(RT)结果的预测和预测价值，并可用于支持治疗修改，但由于生理图像的重复性、肿瘤中图像参数的异质性以及缺乏支持治疗修改的工具等问题，临床应用具有挑战性。我们假设，考虑到肿瘤的潜在异质性及其对治疗的反应，生理成像可以通过稳健和定量地识别具有比肿瘤整体成像特征更好的预测价值的子体积来最好地用于个体化RT，并且可能是局部失败的部位。这些子体积可以作为辐射剂量再分布的空间指南，例如焦点增强，以改善结果。我们开发和研究了模糊逻辑模式识别技术，用于从不同患者和多个时间点的肿瘤血容量(BV)的不同分布中识别头颈(HN)癌的这些亚体积。基于我们的发现，在治疗前HN肿瘤的大的低灌注子体积并且在放化疗的早期过程中持续存在，比起肿瘤中平均BV的变化，我们有可能更好地预测局部失败，我们将进一步发展亚体积定义方法，将其扩展到扩散加权(DW)MR成像，并在一项针对预后不良的HN肿瘤的随机II期临床试验中对其进行评估和验证。我们的目标是：(1)开发定量和自动化的方法，从动态对比增强(DCE)和DW MRI中提取HN肿瘤的重要亚体积，用于预测局部和区域性失败；(2)使用重测数据，前瞻性评估由我们的方法提取的亚体积强度和定义的变异性和重复性；以及(3)在一项预后较差的HN癌的放射剂量增加的随机第二阶段试验中，前瞻性评估和验证该方法产生的预测局部和区域性失败的亚体积。影响：除非临床能够有效和定量地获取和分析这些图像，并以准确的方式使用预测结果，否则这些强大的成像技术在肿瘤RT中产生有意义的差异的好处可能会完全被掩盖。本项目致力于实现这些目标。在该项目中开发的框架可以翻译成其他身体部位以及其他成像模式。通过与一家商业公司合作，我们将使社区中的用户可以使用这种方法。

DESCRIPTION (provided by applicant): Although metabolic and physiological imaging have shown prognostic and predictive value for radiation therapy (RT) outcome and may serve to support therapy modification, clinical utilization is challenging due to issues such as reproducibility of physiological images, heterogeneity of the image parameters in the tumor, and lack of tools to support therapy modification. We hypothesize that, given the underlying heterogeneity of tumors and their response to treatment, physiological imaging can be best utilized for individualizing RT by robustly and quantitatively identifying subvolumes that have a better predictive value than the imaging characteristics of the tumor as a whole, and are the likely sites of local failure. These subvolumes can serve spatial guide for radiation dose redistribution such as focal boosting to improve outcomes. We have developed and investigated fuzzy logic pattern recognition techniques for identifying these subvolumes of head-and-neck (HN) cancer from heterogeneous distributions of tumor blood volume (BV) across patients and over multiple time points. Based on our findings that large poorly-perfused subvolumes of HN tumors before treatment that persist during the early course of chemo-RT have the potential to predict local failure better than the change in the mean BV in the tumor, we will further develop the subvolume definition method, extend it to diffusion- weighted (DW) MR imaging, and evaluate and validate it in a randomized phase II clinical trial of poor- prognosis HN cancers. Our aims are: (1) Develop quantitative and automated methods to extract significant subvolumes of HN tumors from dynamic contrast enhanced (DCE) and DW MRI for prediction of local and regional failure; (2) Prospectively assess variability and reproducibility of the subvolume intensity and definition extracted by our methods using test-retest data; and (3) Prospectively evaluate and validate that the method yields subvolumes predictive of local-regional failure in a randomized phase II trial of radiation dose boosting for poor-prognosis HN cancer. Impact: Unless clinics can acquire and analyze these images effectively and quantitatively, and use predictive results in an accurate fashion, the benefit of these powerful imaging techniques to make a meaningful difference in tumor RT may be completely obscured. This project strives to achieve these goals. The framework developed in the project can be translated into other body sites as well as other imaging modalities. By partnering with a commercial company, we will make this method available for the users in the community.