基于数字化乳腺X线影像组学预测浸润性乳腺癌腋窝淋巴结转移的多中心研究
Radiomics nomogram based on digital mammography for predicting axillary lymph node metastasis of in-vasive breast cancer:a multicenter study
谢玉海 1马培旗 2王小雷 1韩剑剑 3马文俊 1曹雪花 1张宁宁 1杨杨 1胡东1
作者信息
- 1. 236600 安徽,太和县人民医院/皖南医学院附属太和医院放射影像科
- 2. 236000 安徽,安徽省阜阳市人民医院放射影像科
- 3. 241000 安徽,皖南医学院第一附属医院/弋矶山医院放射科
- 折叠
摘要
目的:探讨基于多中心数字化乳腺X线影像组学预测浸润性乳腺癌腋窝淋巴结转移的临床应用价值.方法:回顾性搜集 728 例经病理证实的浸润性乳腺癌患者,按照 7:3 的比例将皖南医学院第一附属医院弋矶山医院 413 例浸润性乳腺癌患者随机拆分为训练组 289 例(淋巴结转移阴性 197 例,淋巴结转移阳性 92 例)和验证组 124 例(淋巴结转移阴性 85 例,淋巴结转移阳性 39 例),将阜阳市人民医院和太和县人民医院浸润性乳腺癌患者共计 315 例(淋巴结转移阴性 221 例,淋巴结转移阳性 94 例)作为外部测试组.对比分析双乳内外斜位(MLO)和头尾位(CC)图像,选取病变面积较大的数字化乳腺X线图像使用深睿医疗多模态科研平台进行图像分割及影像组学特征提取,采用特征间线性相关性分析与最小绝对收缩和选择算法(LASSO)对组学特征进行降维并使用支持向量机(SVM)分类器构建影像组学预测模型.采用受试者工作特征(ROC)曲线和决策曲线分析(DCA)评价模型性能.结果:最终筛选出 8 个影像组学特征构建预测模型用于预测浸润性乳腺癌腋窝淋巴结转移,该模型的预测效能在训练组、验证组和外部测试组分别为 0.807、0.790 和 0.753,敏感度、特异度分别为 84.8%和 61.4%、79.5%和 69.4%、44.7%和 92.8%.决策曲线证实了该模型的临床实用性.结论:基于数字化乳腺 X线影像组学对浸润性乳腺癌腋窝淋巴结转移的预测具有较高效能,对患者制定个性化的治疗方案和预后评估有着重要的临床应用价值.
Abstract
Obiective:To investigate the clinical value of radiomics nomogram based on multi-center digital mammography in predicting axillary lymph node metastasis of invasive breast cancer.Methods:A total of 728 patients with pathologically proven invasive breast cancer were retrospectively collected.413 patients with invasive breast cancer from Yiji Mountain Hospital,the First Affiliated Hospital of Wannan Medical College,were randomly divided into a training group of 289 patients(197 cases with negative lymph node metastasis and 92 cases with positive lymph node metastasis)and a validation group of 124 patients(85 cases with negative lymph node metastasis and 39 cases with posi-tive lymph node metastasis)in the ratio of 7:3.A total of 315 patients with invasive breast cancer(221 cases with negative lymph node metastasis and 94 cases with positive lymph node metastasis)from Fuyang People's Hospital and Taihe County People's Hospital were selected as the external test groups.The images from medio-lateral oblique(MLO)and cranio-caudal(CC)views of two breasts were compared and analyzed,and the digital mammogram images with larger lesion area were selected for image segmentation and radiomics feature extraction using the DeepMed multimodal research plat-form.The dimension of radiomics features was reduced by Linear correlation analysis between features and Least absolute shrinkage and selection operator(LASSO)and a prediction model was constructed by a support vector machine(SVM)classifier.The performance of the model was evaluated by Receiv-er Operating Characteristic(ROC)and Decision Curve Analysis(DCA).Results:Eight radiomics fea-tures were finally selected to construct a model for predicting axillary lymph node metastasis in inva-sive breast cancer.The prediction performance of the model was 0.807,0.790 and 0.753 in the training group,validation group and external test group,respectively,with sensitivities and specificities of 84.8%and 61.4%,79.5%and 69.4%,44.7%and 92.8%,respectively.The decision curve confirmed the clinical practicability of the model.Conclusion:Radiomics based on digital mammography has high efficiency in predicting axillary lymph nodes metastasis in invasive breast cancer and has important clinical application value for the formulation of individualized treatment plans and prognosis assess-ment for patients.
关键词
乳腺癌/数字乳腺X线摄影/腋窝淋巴结转移/影像组学Key words
Breast cancer/Digital mammography/Axillary lymph node metastasis/Radiomics引用本文复制引用
基金项目
皖南医学院科研项目(JXYY202139)
北京医学奖励基金会睿影科研基金(YXJL-2022-0105-0116)
出版年
2024
NETL
NSTL
维普
万方数据