何迪梁,赵建新,潘妮妮,施柳言,熊恋秋,马丽丽,赵致平,赵莲萍,黄刚.基于增强CT影像组学及基因组学模型预测卵巢浆液性囊腺癌预后[J].中国医学影像技术,2024,40(5):745~751
基于增强CT影像组学及基因组学模型预测卵巢浆液性囊腺癌预后
Models based on contrast enhanced CT radiomics and imaging genomics for predicting prognosis of ovarian serous cystadenocarcinoma
投稿时间:2023-11-12  修订日期:2024-01-02
DOI:10.13929/j.issn.1003-3289.2024.05.024
中文关键词:  卵巢  囊腺癌,浆液  预后  影像基因组学  体层摄影术, X线计算机
英文关键词:ovary  cystadenocarcinoma, serous  prognosis  imaging genomics  tomography, X-ray computed
基金项目:
作者单位E-mail
何迪梁 甘肃中医药大学第一临床医学院, 甘肃 兰州 730000  
赵建新 甘肃中医药大学第一临床医学院, 甘肃 兰州 730000  
潘妮妮 甘肃中医药大学第一临床医学院, 甘肃 兰州 730000  
施柳言 甘肃中医药大学第一临床医学院, 甘肃 兰州 730000  
熊恋秋 甘肃中医药大学第一临床医学院, 甘肃 兰州 730000  
马丽丽 甘肃中医药大学第一临床医学院, 甘肃 兰州 730000  
赵致平 甘肃省武威肿瘤医院放射科, 甘肃 武威 733000  
赵莲萍 甘肃省人民医院放射科, 甘肃 兰州 730000  
黄刚 甘肃省人民医院放射科, 甘肃 兰州 730000 keen0999@163.com 
摘要点击次数: 152
全文下载次数: 120
中文摘要:
      目的 探讨基于治疗增强前动脉期CT构建的影像组学模型和影像-基因组学模型预测卵巢浆液性囊腺癌(OSC)患者预后的价值。方法 回顾性收集2个中心及癌症影像档案馆(TCIA)共110例OSC增强动脉期CT图像,提取其影像组学特征,建立影像组学Cox回归预后预测模型;于癌症基因组图谱(TCGA)数据库获取399例OSC(TCGA-OV)转录组数据,以Pearson相关系数筛选与预测模型纳入特征相关的基因,进行关联基因富集分析。以Cox回归及蛋白质互作网络(PPI)对57例同时具有完整影像学及转录组学资料的OSC患者(TCGA-TCGA-OV)筛选连接度最高的预后枢纽基因,建立影像-基因组学模型。观察上述模型预测OSC预后的效能。结果 基于5个OSC预后相关影像组学特征的建立影像组学模型在相应训练集和测试集的C指数(C-index)分别为0.782和0.735;纳入30个预后枢纽基因建立的影像-基因组学模型在相应训练集和测试集的C-index分别为0.673和0.659。以上述模型分层后,不同层次患者间生存率差异有统计学意义(P均<0.05)。与影像组学模型相关联的1 135个mRNA基因涉及细胞黏附等生物学行为和PI3K-Akt、细胞外基质受体互作通路及1型糖尿病通路等信号通路。结论 影像组学模型能较好地预测OSC预后;分析OSC mRNA生物信息学可为影像组学模型提供生物学可解释性。
英文摘要:
      Objective To explore the value of model established with radiomics features based on contrast enhanced arterial phase CT and model with radiogenomics for predicting prognosis of ovarian serous cystadenocarcinoma (OSC). Methods Enhanced arterial phase CT images of 110 OSC patients were retrospectively collected from 2 centers and The Cancer Imaging Archive (TCIA) database. The radiomics features were extracted, among those related to prognosis were selected to establish a radiomics Cox regression model. Genes data of 399 OSC patients were obtained from The Cancer Genome Atlas (TCGA) database, and genes related to the radiomics features included in the above radiomics model were identified with high Pearson correlation coefficient, and then enrichment gene analyses were performed. For 57 OSC cases with complete enhanced CT and gene data, the hub genes which had the highest connectivity with radiomics prognosis predicting model were detected using Cox regression and protein-protein interaction (PPI). Furthermore, a radiogenomics prognosis predicting model was established with the hub genes. The efficiencies of these 2 models for predicting prognosis of OSC patients were analyzed. Results Finally, the radiomics model included 5 OSC prognosis-related radiomics features, with C-index of 0.782 and 0.735 in corresponding training and test set, respectively. Meanwhile, the radiogenomics model included 30 prognostic hub genes, with C-index of 0.673 and 0.659 in corresponding training and test set, respectively. The survival rates of patients with better predicted prognosis according to radiomics model and radiogenomics model were both higher compared with the others (both P<0.05). Totally 1 135 mRNA genes were found being associated with radiomics model, including biological behaviors such as cell adhesion, and signaling pathways such as PI3K-Akt, extracellular matrix receptor interaction pathway and type 1 diabetes pathway. Conclusion The radiomics model was effective for predicting prognosis of OSC patients.Analysis of mRNA bioinformatics in OSC patients might provide biological interpretations for the radiomics model.
查看全文  查看/发表评论  下载PDF阅读器