| 郭艺,文娣娣,刘会佳,逯慧珍,郭敏,陈安琪,宦怡,魏梦绮.基于MRI特征构建模型预测乳腺癌Ki-67表达水平[J].中国介入影像与治疗学,2024,21(4):220-223 |
| 基于MRI特征构建模型预测乳腺癌Ki-67表达水平 |
| Models based on MRI features for predicting Ki-67 expression level of breast cancer |
| 投稿时间：2024-01-29 修订日期：2024-02-28 |
| DOI：10.13929/j.issn.1672-8475.2024.04.007 |
| 中文关键词: 乳腺肿瘤 磁共振成像 Ki-67抗原 |
| 英文关键词:breast neoplasms magnetic resonance imaging Ki-67 antigen |
| 基金项目:陕西省重点研发一般项目（2019SF-182） |
|
| 摘要点击次数: 154 |
| 全文下载次数: 134 |
| 中文摘要: |
| 目的 观察基于MRI特征构建的模型预测乳腺癌Ki-67表达水平的价值。方法 回顾性分析62例接受MR检查的女性乳腺癌患者，根据Ki-67表达水平将其分为低表达组（n=8）与高表达组（n=54）；比较组间计数资料，构建logistic回归模型，预测乳腺癌Ki-67表达水平；以线性相关分析评估患者年龄、乳腺癌病灶最大径及表观弥散系数（ADC）值与Ki-67表达水平的相关性，构建多元线性模型；绘制受试者工作特征曲线，以曲线下面积（AUC）评估各模型的预测效能。结果 乳腺癌病灶位置及分叶征均为预测其Ki-67表达水平的独立影响因素（P均＜0.05），据此构建单独及联合logistic回归模型，包括回归模型位置、回归模型分叶及回归模型位置+分叶。Ki-67表达水平与乳腺癌病灶最大径呈正相关（r=0.347，P＜0.05），与其ADC值呈负相关（r=－0.300，P＜0.05），而与患者年龄无明显相关（r=－0.048，P＞0.05）；故基于病灶最大径及ADC值构建线性模型。回归模型位置+分叶预测乳腺癌Ki-67表达水平的效能（AUC=0.903）高于回归模型位置及回归模型分叶（AUC=0.743、0.817）而与线性模型（AUC=0.852）相当。结论 基于MRI特征构建的模型可有效预测乳腺癌Ki-67表达水平；其中，回归模型位置+分叶及线性模型的预测效能较高。 |
| 英文摘要: |
| Objective To observe the value of models based on MRI features for predicting Ki-67 expression level of breast cancer. Methods Data of 62 female breast cancer patients who underwent MR examinations were retrospectively analyzed. The patients were divided into low-expression group （n=8） and high-expression group （n=54） according to the expression level of Ki-67. Enumeration data were compared between groups, and logistic regression models were constructed to predict expression level of Ki-67 in breast cancer. The correlations of patients’ age, the maximum diameter and apparent diffusion coefficient （ADC） values with Ki-67 expression levels of breast cancers were explored using linear correlation analysis, then a multivariate linear model was constructed. Receiver operating characteristic curves were drawn, and the areas under the curves （AUC） were calculated to evaluate the prediction efficacy of each model. Results Both location and lobulation sign were independent impact factors for predicting expression level of Ki-67 in breast cancer （both P＜0.05）. Then logistic regression models were constructed using the above factors alone and in combination, including regression modellocation, regression modellobulation and regression modellocation+lobulation. Ki-67 expression levels were positively correlated with the maximum diameters （r=0.347, P＜0.05）, negatively correlated with ADC values of breast cancer （r=－0.300, P＜0.05）, but not obviously correlated with patients’age （r=－0.048, P＞0.05）. A linear model was constructed based on the maximum diameter and ADC value of breast cancer. The efficacy of regression modellocation+lobulation for predicting Ki-67 expression level in breast cancer （AUC=0.903） was higher than those of regression modellocation and regression modellocation （AUC=0.743, 0.817） while was similar to linear model （AUC=0.852）. Conclusion Models based on MRI features could effectively predict Ki-67 expression level in breast cancers, and modellocation+lobulation and linear model had the higher prediction efficacy. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
