目的探讨MR影像组学模型鉴别HER2低表达与HER2阳性乳腺癌的应用价值。方法回顾性分析我院2018年1月至12月确诊的浸润性乳腺癌233例,其中HER2阳性乳腺癌103例,HER2低表达乳腺癌130例,按8:2随机拆分为训练集186例及测试集47例。基于MR第2...目的探讨MR影像组学模型鉴别HER2低表达与HER2阳性乳腺癌的应用价值。方法回顾性分析我院2018年1月至12月确诊的浸润性乳腺癌233例,其中HER2阳性乳腺癌103例,HER2低表达乳腺癌130例,按8:2随机拆分为训练集186例及测试集47例。基于MR第2期增强图像提取组学特征,数据经过归一化,降维,筛选特征,构建逻辑回归机器学习模型,并于测试集中验证及评估其诊断效能。结果训练集中AUC为0.87,准确率为0.80,敏感性0.89,特异性0.72,PPV0.72,NPV0.89,测试集中AUC值为0.77,准确率0.77,敏感度0.76,特异性0.77,PPV0.73,NPV0.80,基于DCE-MR影像组学的预测模型不仅具有较好的诊断效能,还具有良好的稳定性。结论基于DCE-MR影像组学预测模型鉴别HE R2低表达与H E R2阳性乳腺癌具有较好的诊断效能,有望为后期临床制定精准化及个性化治疗决策提供参考依据。展开更多
Background:In light of the significant clinical benefits of antibody-drug conjugates in clinical trials,the human epidermal growth factor receptor 2(HER2)-low category in breast cancers has gained increasing attention...Background:In light of the significant clinical benefits of antibody-drug conjugates in clinical trials,the human epidermal growth factor receptor 2(HER2)-low category in breast cancers has gained increasing attention.Therefore,we studied the clinicopathological characteristics of Chinese patients with hormone receptor(HR)-positive/HER2-low early-stage breast cancer and developed a recurrence risk prediction model.Methods:Female patients with HR-positive/HER2-low early-stage breast cancer treated in 29 hospitals of the Chinese Society of Breast Surgery(CSBrS)from Jan 2015 to Dec 2016 were enrolled.Their clinicopathological data and prognostic information were collected,and machine learning methods were used to analyze the prognostic factors.Results:In total,25,096 patients were diagnosed with breast cancer in 29 hospitals of CSBrS from Jan 2015 to Dec 2016,and clinicopathological data for 6486 patients with HER2-low early-stage breast cancer were collected.Among them,5629 patients(86.79%)were HR-positive.The median follow-up time was 57 months(4,76 months);the 5-year disease-free survival(DFS)rate was 92.7%,and the 5-year overall survival(OS)rate was 97.7%.In total,412 cases(7.31%)of metastasis were observed,and 124(2.20%)patients died.Multivariate Cox regression analysis revealed that T stage,N stage,lymphovascular thrombosis,Ki-67 index,and prognostic stage were associated with recurrence and metastasis(P<0.05).A recurrence risk prediction model was established using the random forest method and exhibited a sensitivity of 81.1%,specificity of 71.7%,positive predictive value of 74.1%,and negative predictive value of 79.2%.Conclusion:Most of patients with HER2-low early-stage breast cancer were HR-positive,and patients had favorable outcome;tumor N stage,lymphovascular thrombosis,Ki-67 index,and tumor prognostic stage were prognostic factors.The HR-positive/HER2-low early-stage breast cancer recurrence prediction model established based on the random forest method has a good reference value for predicting 5-year 展开更多
文摘目的探讨MR影像组学模型鉴别HER2低表达与HER2阳性乳腺癌的应用价值。方法回顾性分析我院2018年1月至12月确诊的浸润性乳腺癌233例,其中HER2阳性乳腺癌103例,HER2低表达乳腺癌130例,按8:2随机拆分为训练集186例及测试集47例。基于MR第2期增强图像提取组学特征,数据经过归一化,降维,筛选特征,构建逻辑回归机器学习模型,并于测试集中验证及评估其诊断效能。结果训练集中AUC为0.87,准确率为0.80,敏感性0.89,特异性0.72,PPV0.72,NPV0.89,测试集中AUC值为0.77,准确率0.77,敏感度0.76,特异性0.77,PPV0.73,NPV0.80,基于DCE-MR影像组学的预测模型不仅具有较好的诊断效能,还具有良好的稳定性。结论基于DCE-MR影像组学预测模型鉴别HE R2低表达与H E R2阳性乳腺癌具有较好的诊断效能,有望为后期临床制定精准化及个性化治疗决策提供参考依据。
基金supported by grants from the Youth Cultivation Fund of Beijing Medical Ward Foundation(No.20180502)Beijing Medical Ward Foundation(Nos.YXJL-2016-0040-0065,YXJL-2020-0941-0736)Chinese junior breast surgeon research award fund(No.2020-CHPASLP-01)
文摘Background:In light of the significant clinical benefits of antibody-drug conjugates in clinical trials,the human epidermal growth factor receptor 2(HER2)-low category in breast cancers has gained increasing attention.Therefore,we studied the clinicopathological characteristics of Chinese patients with hormone receptor(HR)-positive/HER2-low early-stage breast cancer and developed a recurrence risk prediction model.Methods:Female patients with HR-positive/HER2-low early-stage breast cancer treated in 29 hospitals of the Chinese Society of Breast Surgery(CSBrS)from Jan 2015 to Dec 2016 were enrolled.Their clinicopathological data and prognostic information were collected,and machine learning methods were used to analyze the prognostic factors.Results:In total,25,096 patients were diagnosed with breast cancer in 29 hospitals of CSBrS from Jan 2015 to Dec 2016,and clinicopathological data for 6486 patients with HER2-low early-stage breast cancer were collected.Among them,5629 patients(86.79%)were HR-positive.The median follow-up time was 57 months(4,76 months);the 5-year disease-free survival(DFS)rate was 92.7%,and the 5-year overall survival(OS)rate was 97.7%.In total,412 cases(7.31%)of metastasis were observed,and 124(2.20%)patients died.Multivariate Cox regression analysis revealed that T stage,N stage,lymphovascular thrombosis,Ki-67 index,and prognostic stage were associated with recurrence and metastasis(P<0.05).A recurrence risk prediction model was established using the random forest method and exhibited a sensitivity of 81.1%,specificity of 71.7%,positive predictive value of 74.1%,and negative predictive value of 79.2%.Conclusion:Most of patients with HER2-low early-stage breast cancer were HR-positive,and patients had favorable outcome;tumor N stage,lymphovascular thrombosis,Ki-67 index,and tumor prognostic stage were prognostic factors.The HR-positive/HER2-low early-stage breast cancer recurrence prediction model established based on the random forest method has a good reference value for predicting 5-year
