| 樊霞,王健,夏雨薇,刘文亚.基于MRI影像组学预测肝泡型包虫病边缘微血管侵犯[J].中国医学影像技术,2021,37(12):1849~1853 |
| 基于MRI影像组学预测肝泡型包虫病边缘微血管侵犯 |
| Radiomics based on MRI for predicting microvascular invasion at edge of hepatic alveolar echinococcosis |
| 投稿时间：2021-02-02 修订日期：2021-08-17 |
| DOI：10.13929/j.issn.1003-3289.2021.12.021 |
| 中文关键词: 肝疾病,寄生虫性 棘球蚴病,肝 磁共振成像 影像组学 |
| 英文关键词:liver diseases, parasitic echinococcosis, hepatic magnetic resonance imaging radiomics |
| 基金项目:国家自然科学基金（81974263）。 |
|
| 摘要点击次数: 2023 |
| 全文下载次数: 403 |
| 中文摘要: |
| 目的 评估基于MR T2WI影像组学模型预测肝泡型包虫病（HAE）病灶边缘微血管侵犯的价值。方法 回顾性分析89例经术后病理证实的HAE患者，其中32例病灶边缘存在微血管侵犯、57例无侵犯。提取病灶MR T2WI影像组学特征，以方差阈值法和单变量选择法筛选最优特征，以随机森林（RF）、极限梯度增强树（XGBoost）和逻辑回归（LR）三种分类器构建预测HAE病灶边缘微血管侵犯的机器学习（ML）模型。按8：2比例将患者分为训练集（n=70）和测试集（n=19），验证模型的预测效能；绘制受试者工作特征（ROC）曲线，计算其曲线下面积（AUC）。结果 共提取1 409个影像组学特征，经特征降维选出7个最优影像组学特征，并以之构建模型。ROC曲线显示，XGBoost模型在训练集及测试集中的AUC分别为0.96和0.89。结论 基于MR T2WI影像组学XGBoost模型可有效预测HAE病灶边缘微血管侵犯。 |
| 英文摘要: |
| Objective To observe the value of radiomics model based on MR T2WI for predicting microvascular invasion at the edge of hepatic alveolar echinococcosis (HAE). Methods Preoperative MRI data of 89 patients with HAE confirmed by pathology after surgical resection were retrospectively analyzed, including 32 cases with and 57 without microvascular invasion at the lesions' edges. Then radiomics features of lesions were extracted based on T2WI, and both variance selection method and univariate selection method were used for screening the optimal features. Three classifiers, including random forest (RF), extreme gradient enhancement tree (XGBoost) and Logistic regression (LR) were used to construct machine learning (ML) models for predicting microvascular invasion at the edge of HAE lesions. Then 89 patients were divided into training set (n=70) and test set (n=19) at the ratio of 8:2. The corresponding receiver operating characteristic (ROC) curves were drawn, and the areas under the curves (AUC) were calculated, and the predictive performance of each ML model was observed. Results A total of 1 409 radiomic features were extracted, and 7 optimal radiomic features were screened out by feature dimension reduction to construct machine learning models. ROC curve showed that XGBoost model performed well in both training set and testing set, with AUC of 0.96 and 0.89, respectively. Conclusion Radiomics XGBoost ML model based on MR T2WI could effectively predict microvascular invasion at the edge of HAE lesions. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
