| 江长思,罗燕,唐雪,龚静山.基于CT机器学习模型预测肺腺癌气腔播散[J].中国医学影像技术,2020,36(12):1834~1838 |
| 基于CT机器学习模型预测肺腺癌气腔播散 |
| CT-based machine learning model in prediction of spread through air space of lung adenocarcinoma |
| 投稿时间：2019-11-05 修订日期：2020-06-20 |
| DOI：10.13929/j.issn.1003-3289.2020.12.016 |
| 中文关键词: 肺肿瘤 肿瘤转移 体层摄影术,X线计算机 机器学习 |
| 英文关键词:lung neoplasms neoplasm metastasis tomography, X-ray computed machine learning |
| 基金项目: |
|
| 摘要点击次数: 1809 |
| 全文下载次数: 773 |
| 中文摘要: |
| 目的 探讨基于胸部CT建立机器学习模型预测肺腺癌气腔播散（STAS）状态的价值。方法 回顾性分析462例经手术病理证实的原发性肺腺癌，90例STAS阳性（STAS阳性组），372例STAS阴性（STAS阴性组），比较组间一般资料及CT征象差异。按7 ：3比例将患者随机分为训练集（n=323）和验证集（n=139），采用随机森林算法针对差异有统计学意义的变量建立预测肺腺癌STAS的模型，对训练集进行训练，获得平均AUC最大的模型，以之对验证集进行预测，并计算AUC，评价其诊断效能。结果 组间年龄及12个CT征象差异均具有统计学意义（P均<0.05）。以上述13个变量建立的预测模型对训练集进行预测，获得最大AUC为0.80的模型，其预测训练集和验证集肺腺癌STAS的AUC、敏感度及特异度分别0.80、0.78及0.77和0.77、0.78及0.73。结论 基于胸部CT的机器学习模型可有效预测肺腺癌STAS状态。 |
| 英文摘要: |
| Objective To investigate the value of CT-based machine learning model in prediction of spread through air space (STAS) in lung adenocarcinoma. Methods Imaging and clinical data of 462 lung adenocarcinoma patients confirmed by surgical pathology were retrospectively analyzed, including 90 STAS-positive patients (STAS positive group) and 372 STAS-negative patients (STAS negative group). General information and CT features were observed and compared between groups. All patients were randomly divided into training set (n=323) and validation set (n=139) according to the ratio of 7 : 3. For variables being statistically different between groups, random forest was used to establish models for predicting STAS in lung adenocarcinoma, and the training set applied to obtain model with the largest average AUC. Then the model was used to predict STAS status of validation set, its AUC was calculated, and the corresponding diagnostic efficacy was analyzed. Results Statistical differences of patient's age and 12 CT features were found between groups (all P<0.05). The model with the largest AUC (0.80) was obtained by using the predicted model established with the above 13 variables in training set. The AUC, sensitivity, and specificity of the model for predicting STAS status of lung adenocarcinoma in the training set and the validation set was 0.80, 0.78, 0.77 and 0.77, 0.78, 0.73, respectively. Conclusion CT-based machine learning model could effectively predict the status of STAS of lung adenocarcinoma. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
