魏坤杰,邵硕,郑宁,崔景景,苑子茵,刘诗晗.基于多期相动态对比增强磁共振影像组学在预测浸润性乳腺癌前哨淋巴结转移中的价值[J].济宁医学院学报,2023,46(1):10-15,19 |
基于多期相动态对比增强磁共振影像组学在预测浸润性乳腺癌前哨淋巴结转移中的价值 |
The preoperative prediction value based on multiple phase radiomics models of DCE-MRI in sentinel lymph node metastasis of invasive breast cancer |
投稿时间:2022-03-09 |
DOI:10.3969/j.issn.1000-9760.2023.01.003 |
中文关键词: 乳腺癌;前哨淋巴结转移;影像组学;动态对比增强磁共振成像 |
英文关键词: Breast cancer;Sentinel lymph node metastasis;Radiomics;Dynamic enhanced magnetic resonance imaging |
基金项目: |
作者 | 单位 | E-mail | 魏坤杰 | 济宁医学院临床医学院, 济宁 272013 济宁市第一人民医院, 济宁 272011 | | 邵硕 | 济宁市第一人民医院, 济宁 272011 | | 郑宁 | 济宁市第一人民医院, 济宁 272011 | zhengning_369@163.com | 崔景景 | 联影智能医疗科技(北京)有限公司, 北京 100094 | | 苑子茵 | 济宁医学院临床医学院, 济宁 272013 济宁市第一人民医院, 济宁 272011 | | 刘诗晗 | 济宁医学院临床医学院, 济宁 272013 济宁市第一人民医院, 济宁 272011 | |
|
摘要点击次数: 1328 |
全文下载次数: 714 |
中文摘要: |
目的 探讨基于多期相动态对比增强磁共振(dynamic contrast-enhanced magnetic resonance imaging,DCE-MRI)影像组学在预测浸润性乳腺癌前哨淋巴结(sentinel lymph node,SLN)转移中的价值。方法 回顾性收集2018年11月至2021年9月在济宁市第一人民医院术前接受乳腺DCE-MRI检查且经病理证实的150名浸润性乳腺癌患者的临床、病理及MRI资料,其中,SLN转移阳性者61名,阴性者89名,并将其以8∶2的比例随机划分为训练集(n=120)与测试集(n=30)。在乳腺DCE-MRI(增强早期、增强峰值期及增强末期)剪影图像上进行手动逐层勾画感兴趣区(region of interest, ROI),获得三维容积感兴趣区域(volume of interest,VOI),再对各期进行提取影像组学特征,使用Z分数(Z-Score)归一化对特征进行归一化处理,然后再使用Select K Best和最小绝对收缩与选择算法(least absolute shrinkage and selection operator,LASSO)筛选出最优特征,并构建logistic回归(logistic regression,LR)模型。绘制受试者工作特征(receiver operating characteristic,ROC)曲线及曲线下面积(area under the curve,AUC)。运用ROC曲线与决策曲线分析(decision curve analysis,DCA)对模型进行评价。结果 分别从增强早期、增强峰值期、增强末期及三期联合期相的图像中得到了10、10、10及11个最优特征,通过LR共构建4个预测模型。在训练集中,4个模型的AUC值分别为0.859、0.801、0.768、0.834。在测试集中,4个模型AUC值分别为0.843、0.806、0.806、0.866。DCA显示联合期相模型表现出了较高的净收益。结论 DCE-MRI增强早期、增强峰值期及增强末期影像组学模型在预测浸润性乳腺癌SLN转移中均具有较好的预测效能,且测试集中联合期相的效能略高于单独期相。 |
英文摘要: |
Objective To investigate the value of radiomics models based on multiple phase dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in predicting sentinel lymph node metastasis of invasive breast cancer.Methods The clinical,pathological and MRI data of 150 patients who received breast DCE-MRI examination before operation and were confirmed with invasive breast cancer by pathology in Jining First People’s Hospital from Nov.2018 to Sept.2021 were retrospectively collected.The average age of the patients was (50.28 ±9.58) years old.Among them,61 were positive for sentinel lymph node metastasis and 89 negative.They were randomly divided into training set (n=120) and test set (n=30) at a ratio of 8∶2.The region of interest (ROI) was manually delineated layer by layer on the breast DCE-MRI (enhanced early,peak and later phase) subtraction images to obtain the three-dimensional volume of interest (VOI).Then,radiomics features were extracted for each phase.The Z-score normalization method (Z-Score) was used to eliminate and normalize the features.Then Select K Best and least absolute shrinkage and selection operator (LASSO) were used to select the optimal features,and logistic regression (LR) was used to construct the models.Receiver operating characteristic (ROC) curve and area under the curve (AUC) were drawn.The models was evaluated by ROC curve and decision curve analysis.Results 10,10,10 and 11 optimal features were obtained from the images of enhanced early,peak,later and combined phases,respectively.Four prediction models were constructed by LR.In the training set,the AUC values of four prediction models were 0.859,0.801,0.768 and 0.834 respectively.In the test set,the AUC values of four prediction models were 0.843,0.806,0.806 and 0.866 respectively.DCA showed the combined phase model required good net benefit.Conclusion DCE-MRI at enhanced early,peak and later phase has a good predictive value in predicting sentinel lymph node metastasis of invasive breast cancer.Combined phases can improve the prediction efficiency of the model. |
查看全文
查看/发表评论 下载PDF阅读器 |
|
|
|