| 引用本文： |
-
唐玉洪,郑嘉祺,王 焕,等.不同影像组学标签诊断早期肺腺癌侵袭性的比较研究[J].同济大学学报（医学版）,2019,40(5):585-591. [点击复制]
- TANG Yu-hong,ZHENG Jia-qi,WANG Huan,et al.Association between multi-view radiomics signature and invasion of lung adenocarcinoma[J].同济大学学报（医学版）,2019,40(5):585-591. [点击复制]
|
|
| 摘要: |
| 目的 探讨多视图下影像组学标签在肺腺癌侵袭性诊断的价值。方法 回顾性分析2013年11月—2018年9月经手术病理证实的220例肺腺癌患者的CT图像。从矢状面、冠状面和水平面3个视图分别提取了163个影像组学特征用于构建肺腺癌侵袭性影像组学标签。结果 不同视图下，69个影像组学特征差异具有统计学意义（P值均<0.05）。同一视图下，梯度提升决策树模型构建的影像组学标签在全集和训练集的AUC面积与LASSO模型、朴素贝叶斯模型构建的影像组学标签的AUC面积差异具有统计学意义（P<0.001,P=0.006,P=0.049,P=0.013）。结论 相较于LASSO和朴素贝叶斯模型，梯度提升决策树构建的影像组学标签诊断效能更好。梯度提升决策树联合矢状面构建的影像组学标签诊断效能高于冠状面构建的影像学组标签。 |
| 关键词: 磨玻璃结节 影像组学 多视图 梯度提升决策树 最小绝对收缩和选择算子 朴素贝叶斯 |
| DOI：10.16118/j.1008-0392.2019.05.010 |
| 投稿时间：2019-01-24 |
| 基金项目:国家自然科学基金(81872718)；上海市卫生与健康委员会项目(201840041) |
|
| Association between multi-view radiomics signature and invasion of lung adenocarcinoma |
| TANG Yu-hong,ZHENG Jia-qi,WANG Huan,WANG Bin,SUN Xi-wen,AI Zi-sheng |
| (Dept. of Medical Statistics, Tongji University School of Medicine, Shanghai 200092, China;Dept. of Radiology, Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai 200433, China) |
| Abstract: |
| Objective To investigate the association between multi-view radiomics features and the invasion of lung adenocarcinoma. Methods The clinical data and CT images of 220 pathologically confirmed lung adenocarcinomas were retrospectively reviewed. Total 163 radiomics features were exacted from sagittal plane, coronal plane and transverse plane of CT images to build radiomics signature scores. The association of radiomics signature and invasion of lung adenocarcinoma was analyzed. Results The 69 radiomics features were statistically significant among different views. In coronal plane, the radiomics signature score built by GBDT demonstrated best discrimination with AUC in all dataset and train dataset than other radiomics signature scores built by LASSO and Nave Bayes. Compared with coronal plane, the radiomics signature score built by GBDT in sagittal plane showed better diagnosis performance（P＜0.001,P=0.006,P=0.049,P=0.013）. Conclusion The AUC of radiomics signature score built by GBDT in different views are statistically significant. And the radiomics signature score built by GBDT in sagittal plane shows the best diagnosis performance in lung adenocarcinoma. |
| Key words: ground glass nodule radiomics multi-view gradient boosting decision tree least absolute shrinkage and selection operator naive Bayes |
