| 陈基明,朱浩雨,高静,葛亚琼,王敏红,李颖,吴莉莉.基于临床病理及常规和功能MRI影像组学模型预测乳腺癌腋窝淋巴结转移[J].中国医学影像技术,2021,37(6):885~890 |
| 基于临床病理及常规和功能MRI影像组学模型预测乳腺癌腋窝淋巴结转移 |
| Radiomics models based on clinical-pathology and conventional and functional MRI for predicting lymph node metastases of breast cancer axillary |
| 投稿时间：2020-04-25 修订日期：2021-03-25 |
| DOI：10.13929/j.issn.1003-3289.2021.06.022 |
| 中文关键词: 乳腺肿瘤 淋巴结转移 磁共振成像 影像组学 |
| 英文关键词:breast neoplasms lymphatic metastasis magnetic resonance imaging radiomics |
| 基金项目: |
|
| 摘要点击次数: 3160 |
| 全文下载次数: 719 |
| 中文摘要: |
| 目的 评估基于临床病理及常规和功能MRI（fMRI）影像组学模型预测乳腺癌腋窝淋巴结（ALN）转移的价值。方法 回顾性分析140例浸润性乳腺癌，按7∶3比例将其分为训练集（n=99）和验证集（n=41）。采用多因素Logistic回归分析分别建立基于临床病理及MRI特征的临床模型及各序列图像影像组学、联合序列影像组学以及临床病理及常规和fMRI影像组学的个体化模型，以受试者工作特征（ROC）曲线评价其诊断效能；比较个体化模型与临床模型曲线下面积（AUC）的差异，应用决策曲线分析（DCA）评估模型的临床获益。结果 临床模型预测训练集和验证集ALN转移的AUC分别为0.95和0.88；T2WI、DWI、DCE-MRI模型及联合序列模型在验证集中的AUC分别为0.67、0.71、0.72及0.76。个体化模型在训练集和验证集中的AUC为0.98和0.93，与临床模型差异均无统计学意义（Z=1.56、1.34，P=0.12、0.18）。DCA结果显示阈值>0.25时，个体化模型的净受益高于临床模型。结论 基于临床病理及常规和功能MRI的个体化模型预测乳腺癌ALN转移的效能与临床模型相当，其净受益高于后者，且均优于单一序列模型。 |
| 英文摘要: |
| Objective To investigate the value of radiomics models based on clinical-pathology and conventional and functional MRI (fMRI) in predicting axillary lymph node (ALN) metastases of breast cancer. Methods Data of 140 patients with histologically confirmed breast cancer were retrospectively analyzed. The patients were divided into training set (n=99) and validation set (n=41) at the ratio of 7 ∶3. Multivariable Logistic regression analysis was used to establish the classification models based on clinical-pathology and MRI characteristics, sequences radiomics and combined sequences radiomics, as well as individualized model based on clinical-pathology combined with conventional and fMRI, respectively. The performance of the models were assessed with the receiver operating characteristic (ROC) curve. The area under the curve (AUC) of individualized model and clinical model were compared. The clinical value of models were analyzed using decision curve analysis (DCA). Results AUC of clinical model for predicting ALN metastases in training and validation set was 0.95 and 0.88, respectively, of T2WI, DWI, DCE-MRI and combined model in validation set was 0.67, 0.71, 0.72 and 0.76, respectively. AUC of individualized model was 0.98 and 0.93 in training and validation set, showing no statistical difference with those of clinical model (Z=1.56, 1.34, P=0.12, 0.18). DCA showed that when threshold>0.25, the net benefit of individualized model was superior to that of clinical model. Conclusion The efficacy of individualized model based on clinical-pathology and conventional and functional MRI was similar to that of clinical model in predicting breast cancer ALN metastases, but its clinical benefit was higher than that of the latter, while both of which were superior to single sequence models. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
