基于CT成像的胰腺癌影像基因组学的基础及临床研究

With the progress of imaging equipment, computer graphics/image processing software, and the in-depth understanding of tumor gene research, radiomics have drawn the attention of radiologists because of the advantages of visual and quantitative evaluation, and it has been used for diagnostic and prognostic researches in various tumors. However, few studies have been reported in the most malignant pancreatic carcinoma. Therefore, in our previous experiments, which based on CT venous phase images from 77 patients of pancreatic carcinoma, we preliminarily established a radiomics model and discovered that it can be used to differentiate between early stage and advanced stage pancreatic carcinoma (AUC = 0.759)..Two hundred cases of pancreatic carcinoma will be prospectively collected, including CT multi-phase enhanced images, clinical data, blood samples, pathological tissue specimens and related gene information, using the interactive segmentation method to draw the region of interest, and then extracting and screening high-throughput CT image features to construct the radiomics model. Through the screening of public gene database, and RT-PCR and immunohistochemistry detection of 200 cases of pancreatic carcinoma tissue & blood samples, we aimed to find the key genes in pancreatic carcinoma, to analyze the relationship between the CT images features and the tumor genes, to construct radiogenomics models to diagnosing early stage pancreatic carcinoma and predicting prognosis, and try to predict the genotyping of pancreatic carcinoma by imaging features, which will provide a new way for early stage diagnosis and prognosis evaluation of pancreatic carcinoma.

随着影像设备和计算机图形/图像处理软件的进步，以及对肿瘤基因研究和认识的深入，可视化、定量评估的影像组学方法受到了影像医师的关注，用于各种肿瘤的诊断、预后等方面的研究，但在极具恶性的胰腺癌中研究较少。因而，我们前期实验、基于77例胰腺癌CT静脉期图像，初步构建了影像组学模型，发现可鉴别早期与进展期胰腺癌（AUC=0.759）。.本项目将前瞻性收集200例有CT多期增强图像、临床信息、血液样本、病理组织标本及相关基因信息的胰腺癌资料，采用感兴趣区交互式分割、高通量CT影像特征提取与筛选的方法，构建影像组学模型；通过公共数据库基因筛选、200例胰腺癌组织和血液样本的RT-PCR、免疫组化等基因检测方法寻找胰腺癌的关键基因；分析CT影像特征与肿瘤基因的相互关系，构建胰腺癌早期诊断及预后相关的影像基因组学模型，并尝试通过影像特征预测胰腺癌的基因分型，为胰腺癌早期诊断及预后评估提供新途径。

胰腺癌（pancreatic carcinoma,PC）已成为国内外医疗界需要共同面对的一个重大的诊断和治疗难题，其早期诊断困难、恶性程度高、进展迅速、预后极差，被称为癌中之王。因此寻找胰腺癌的早期诊断及预后标志物，对于胰腺癌的精准诊疗具有十分重要的意义。传统影像学方法对PC的诊断主要在解剖学和形态学层面，难以实时全面反映病变的情况。术前获取肿瘤组织需要通过穿刺活检或胰液细胞学检查，它们具有侵入性、获取组织有限、并发症多、受操作者和肿瘤部位影响大，运用于临床受到一定限制。目前急需一种既便捷又准确高效的诊断方法对PC进行个体化的预测。本项目主要包含两方面的研究：1）基于CT成像的影像组学方法，构建预测模型，实现对患者的个体化预测。包括对胰腺癌CT影像进行特征提取和筛选，临床病理指标的获取和筛选以及影像组学预测模型的构建、验证和优化。可在术前较准确的预测胰腺癌的临床分期和病理分级。本研究的完成为胰腺癌的精准诊疗提供了无创评价的新方法，进一步为实现个体化治疗提供了重要参考。2）通过文献阅读、数据挖掘与标本分析，筛选出与胰腺癌预后相关的基因，并进行生物信息学分析。利用基于与焦亡预后相关基因结合临床信息构建Nomogram模型，实现对胰腺癌患者进行风险分层。本研究在临床上为胰腺癌患者的预后评估具有重要的指导价值，进一步为实现个体化治疗提供了重要参考。本项目为胰腺癌精准诊疗的发展提供了新的思路和方法。