基因-多模态影像联合标志驱动的放射性脑损伤精准预测模型研究

Radiation brain injury (RBI) is one of the most serious complications after radiotherapy in patients with nasopharyngeal carcinoma (NPC). Early prediction and intervention is the key to improving prognosis. The occurrence of RBI has a certain genetic predisposition, and the application of single nucleotide polymorphism (SNP) in predicting and screening the patients susceptible to radiation injury is a hot issue. Previous studies of RBI have simply observed the image changes without clarifying the genetic basis. In order to meet the urgent needs of individualized treatment, it is necessary to predict the incidence risk of RBI by establishing a predictive model based on various risk factors. This project aims to carry out a longitudinal study using multimodality imaging and genes based on our previous studies on the structural and functional integration of NPC patients. We plan to collect the multimodality imaging on different stages before and after radiotherapy and evaluate the polymorphism of radiation injury related genes ATM, P53, XRCC, ERCC, TGFβ and their miRNA expression levels, as well as cognitive functions of NPC patients. The association of imaging markers with genes and cognitive functions will be investigated. Furthermore, in combination with clinical risk factors, radiomics high-throughput feature analysis is to be applied to establish a precise RBI prediction model. This project will not only provide a new target for early clinical intervention and individualized treatment for RBI, but also show scientific evidence for the pathogenesis of RBI.

鼻咽癌（NPC）放疗后放射性脑损伤（RBI）是最严重并发症之一，早期预测并干预是改善预后的核心问题。RBI的发生有一定遗传倾向，将单核苷酸多态性（SNP）用于早期预测并筛选放射损伤易感者是研究热点。当今RBI相关研究仅观察影像变化，并未阐明其遗传学基础。为了满足个体化治疗的迫切需要，有必要综合RBI的各危险因素建立预测模型用以准确预测发病风险。本项目拟在NPC放疗后脑结构与功能等改变的研究基础上，综合多模态影像与基因对NPC患者行纵向研究：收集放疗前后不同时期的多模态影像，分析其相关基因ATM、P53、XRCC、ERCC及TGFβ的多态性及其miRNA的表达水平，评估认知功能，探讨影像标志物与基因及认知功能之间的相关性；综合各种危险因素，采用影像组学高通量特征分析，建立RBI精准预测模型，该项目的实施将为临床早期干预RBI及实行个体化治疗提供新靶点，并为探索RBI发病机理提供科学依据。

鼻咽癌放疗后放射性脑损伤是最严重并发症之一，早期预测并干预是改善预后的核心问题。放射性脑损伤的发生有一定遗传倾向，将单核苷酸多态性用于早期预测并筛选放射损伤易感者是研究热点。当今放射性脑损伤相关研究仅观察影像变化，并未阐明其遗传学基础。为了满足个体化治疗的迫切需要，有必要综合放射性脑损伤的各危险因素建立预测模型用以准确预测发病风险。本项目在鼻咽癌放疗后脑结构与功能等改变的研究基础上，应用多模态MR及多变量模式分析方法，提取了放射性脑损伤的脑影像标志物，我们利用机器学习方法和DTI技术发现放疗后各组与正常对照组均可以互相区分开来，并能达到较高的识别率，放射性脑损伤中最具疾病区分能力的白质区位于双侧颞叶及双侧小脑，并且它是一种全脑白质网络发生异常的疾病。并且应用基于体素的形态学方法发现放疗后脑灰质容积显著减小，且与认知功能减退显著相关。同时我们采用图论的分析方法对Bold-fMRI数据进行了分析，发现鼻咽癌放疗后患者静态和动态功能网络稳定性的破坏、网络效率的降低和功能连接的降低可能是放射性脑损伤的潜在生物标志物。此外，对已采集的2组被试的血液样本进行相关基因检测，初步结果显示，鼻咽癌放疗后患者认知功能障碍的发生与ATM基因多态性相关。上述发现将为临床早期干预放射性脑损伤及实行个体化治疗提供新靶点，并为探索放射性脑损伤发病机理提供科学依据。