| 喻泓清,翟建,刘晨露,肖国庆,李庆祝.基于MRI影像组学预测宫颈癌淋巴血管间隙浸润[J].中国医学影像技术,2022,38(3):421~426 |
| 基于MRI影像组学预测宫颈癌淋巴血管间隙浸润 |
| Radiomics based on MRI for predicting cervical cancer lymph-vascular space invasion |
| 投稿时间：2021-06-30 修订日期：2021-10-17 |
| DOI：10.13929/j.issn.1003-3289.2022.03.023 |
| 中文关键词: 子宫颈肿瘤 磁共振成像 影像组学 淋巴血管间隙浸润 |
| 英文关键词:uterine cervical neoplasms magnetic resonance imaging radiomics lymph-vascular space invasion |
| 基金项目: |
|
| 摘要点击次数: 2171 |
| 全文下载次数: 565 |
| 中文摘要: |
| 目的 评估基于MRI影像组学模型术前预测宫颈癌淋巴血管间隙浸润（LVSI）的价值。方法 回顾性分析123例经病理证实宫颈癌患者，根据病理结果分为LVSI+（n=61）及LVSI-（n=62）。基于T2WI及动脉期对比增强T1WI （CE-T1WI）提取影像组学特征，按7∶3比例将数据分为训练集（n=87）和验证集（n=36），以最大相关最小冗余（mRMR）和最小绝对收缩和选择算子（LASSO）回归对训练集数据进行特征降维，筛选最佳影像组学特征，构建预测宫颈癌LVSI的单一序列影像组学模型（模型T2WI、模型CE-T1WI）；应用多因素logistic回归分析构建基于T2WI+CE-T1WI的影像组学模型（模型T2WI+CE-T1WI）及临床影像组学模型；分别以受试者工作特征（ROC）曲线及决策曲线分析（DCA）评估其预测效能及临床效益。结果 基于T2WI和CE-T1WI各提取1 316个影像组学特征，最终分别获得13、15个最佳特征，用于构建模型T2WI及模型CE-T1WI。共筛选出13个影像组学特征用于构建模型T2WI+CE-T1WI，其在训练集和验证集数据中的曲线下面积（AUC）分别为0.79及0.78；临床影像组学模型在训练集和验证集中的AUC分别为0.88及0.83，与模型T2W1+CE-T1WI差异均无统计学意义（P均>0.05）。DCA提示，阈值取0.08~0.68时，临床影像组学模型的净收益高于模型T2WI+CE-T1WI。结论 基于T2WI+CE-T1WI的临床影像组学模型术前预测宫颈癌LVSI的效能较高。 |
| 英文摘要: |
| Objective To explore the value of radiomics model based on MRI for preoperatively predicting cervical cancer lymph-vascular space invasion (LVSI). Methods Data of 123 patients with cervical cancer confirmed by pathology were retrospectively analyzed. The patients were classified as LVSI+ (n=61) and LVSI- (n=62) according to pathological results. The texture features were extracted based on T2WI and contrast enhanced T1WI (CE-T1WI), and the data were divided into training set (n=87) and validation set (n=36) at the ratio of 7:3. The maximum relevance minimum redundancy (mRMR), the least absolute shrinkage and selection operator (LASSO) regression were used to reduce the dimension of features in training set for selecting the best texture features. The single sequence radiomics models (modelT2WI, modelCE-T1WI) were established to predict cervical cancer LVSI. Multivariate logistic regression analysis was used to establish radiomics models (modelT2WI+CE-T1WI) based on T2WI+CE-T1WI and clinical radiomics model. Receiver operating characteristic (ROC) curve and decision curve analysis (DCA) methods were used to evaluate the predictive efficiency and clinical benefit of the models, respectively. Results Totally 1 316 radiomics features were extracted from T2WI and CE-T1WI, and 13 and 15 best radiomics features were obtained, respectively, then modelT2WI and modelCE-T1WI were constructed. Totally 13 radiomics features were selected based on T2WI and CE-T1WI, and modelT2WI+CE-T1WI was established. The area under the curve (AUC) of modelT2WI+CE-T1WI in training set and validation set was 0.79 and 0.78, while of clinical radiomics model was 0.88 and 0.83, respectively, no significant difference was found between these 2 models (both P>0.05). DCA demonstrated that taken 0.08-0.68 as the threshold, the net benefit of clinical radiomics model was higher than that of modelT2WI+CE-T1WI. Conclusion Clinical radiomics model based on T2WI+CE-T1WI had high diagnostic efficacy in preoperative prediction of cervical cancer LVSI. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
