| 蔡俊辉,段绍峰,袁虎,吕燕,吴雅蔚,许晴,叶靖.机器学习在鉴别肺纯磨玻璃结节浸润性腺癌与非浸润性腺癌中的应用[J].中国医学影像技术,2020,36(3): |
| 机器学习在鉴别肺纯磨玻璃结节浸润性腺癌与非浸润性腺癌中的应用 |
| The application of machine learning in differentiating pulmonary invasive adenocarcinoma from non-invasive adenocarcinoma manifested as pure ground-glass nodule |
| 投稿时间：2019-08-09 修订日期：2020-03-16 |
| DOI： |
| 中文关键词: 影像组学 机器学习 纯磨玻璃结节 肺腺癌 CT |
| 英文关键词:Radiomic Machine learning Pure ground-glass nodule Lung adenocarcinoma CT |
| 基金项目:无 |
|
| 摘要点击次数: 3169 |
| 全文下载次数: 654 |
| 中文摘要: |
| 目的 研究基于影像组学特征构建的机器学习模型对鉴别肺纯磨玻璃结节浸润性腺癌与非浸润性腺癌的价值。方法 回顾性分析经手术病理证实的87例CT表现为纯磨玻璃结节的肺腺癌，其中浸润性腺癌32例，非浸润性腺癌55例（包括原位癌17例，微浸润性腺癌38例）。应用ITK-SNAP软件勾化感兴趣区（ROI），A.K.软件（Artificial Intelligent Kit，GE health）进行影像组学特征提取。采用单因素方差分析、秩和检验及t检验筛选有意义的特征参数，Spearman相关性分析和Lasso回归分析进行特征降维。选取降维后的特征参数分别构建支持向量机（SVM）、随机森林（RF）、逻辑回归（LR）三种机器学习模型，采用十折交叉验证法得到最优模型，并绘制ROC曲线用于评价三种模型的性能。结果 共提取396个影像组学特征，通过特征筛选后最终得到19个影像组学特征。SVM、RF、logistics回归三种机器学习模型可以有效鉴别两组病变，准确率（Accuracy）依次为93.30%、86.70%、83.30%，曲线下面积（AUC）分别为0.94、0.92、0.83。结论 基于影像组学特征构建的机器学习模型有较好的分类性能，说明利用机器学习的方法可以在术前有效的鉴别浸润性腺癌与非浸润性腺癌。 |
| 英文摘要: |
| Objective To investigate the value of machine learning model based on radiomic features in differentiating pulmonary invasive adenocarcinoma from non-invasive adenocarcinoma manifested as pure ground-glass nodule (pGGN). Methods Eighty-seven lung adenocarcinomas presented as pGGN in CT were analyzed retrospectively, there were 32 cases of invasive adenocarcinoma (IAC) and 55 cases of non-invasive adenocarcinoma (17 adenocarcinomas in-situ and 38 minimally invasive adenocarcinomas). The software ITK-SNAP was used to draw the region of interest (ROI), and the radiomic features were extracted using A.K. (Artificial Intelligent Kit, GE health) analysis software. One-way analysis of variance, Mann-Whitney U test and t-test were used to screen the significant characteristic parameters. Then feature dimensionality reduction was conducted with Spearman analysis and LASSO regression. The final feature parameters were selected to construct three machine learning models including Support Vector Machine (SVM), Random Forest (RF) and Logistics Regression (LR). 10-fold Cross Validation was used to get the optimal model and the ROC curve was drawn to evaluate the performance of three models. Results A total of 396 radiomic features were extracted, and 19 features were finally obtained after feature screening. The three machine learning models of SVM, RF, and logistics regression can effectively distinguish the two groups of lesions with the accuracies of 93.30%, 86.70%, and 83.30% respectively, and AUC of 0.94, 0.92 and 0.83 respectively. Conclusion The machine learning model based on radiomic features has good classified performance, which indicates that the method of machine learning can effectively distinguish invasive adenocarcinoma from non-invasive adenocarcinoma preoperatively. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
