Department of Ultrasound, The First Affiliated Hospital of Nanjing Medical University, No. 300 Guangzhou Road, Nanjing, 210029, China.
Department of Radiology, The First Affiliated Hospital of Nanjing Medical University, No. 300 Guangzhou Road, Nanjing, 210029, China.
Eur J Radiol. 2021 Feb;135:109512. doi: 10.1016/j.ejrad.2020.109512. Epub 2020 Dec 31.
To develop a combined nomogram by incorporating the Memorial Sloan Kettering Cancer Center (MSKCC) nomogram and ultrasound (US)-based radiomics score (Radscore) for predicting sentinel lymph node (SLN) metastasis in invasive breast cancer.
This retrospective study was approved by the ethics committee of our institution, and written informed consent was waived. A total of 452 patients with invasive breast cancer who received SLN Biopsy in a single center were included between January 2016 and December 2019. The patients were divided into a training set (n = 318) and a validation set (n = 134). A total of 1216 features were extracted from the regions of interest (ROIs) of the tumors on conventional ultrasound. The maximum relevance minimum redundancy (mRMR) and the least absolute shrinkage and selection operator (LASSO) algorithm were used to build the Radscore. Afterward, the diagnostic performance was assessed and validated. Comparison of receiver operating characteristic (ROC) curves and decision curve analysis (DCA) were performed to evaluate the incremental value of the combined model.
Obtained from 18 features, the Radscore indicated a favorable discriminatory capability in the training set with an area under the curve (AUC) of 0.834, whereas a value of 0.770 was observed in the validation set. The AUC of the combined model was 0.901 (95 % confidence interval (95 % CI): 0.865-0.938) in the training set and 0.833 (95 % CI: 0.788-0.878) in the validation set. Both of them were superior to MSKCC or imaging Radscore alone (P < 0.05). DCA demonstrated that the combined model was superior to the others in terms of clinical practicability.
Preoperative US-based Radscore can improve the accuracy of clinical MSKCC nomogram for SLN metastasis prediction in breast cancer.
通过纳入 Memorial Sloan Kettering Cancer Center（MSKCC）列线图和基于超声的放射组学评分（Radscore），建立一种联合列线图，以预测浸润性乳腺癌前哨淋巴结（SLN）转移。
本回顾性研究经我院伦理委员会批准，并豁免了书面知情同意。纳入 2016 年 1 月至 2019 年 12 月在单一中心接受 SLN 活检的 452 例浸润性乳腺癌患者。患者被分为训练集（n=318）和验证集（n=134）。从肿瘤的感兴趣区域（ROI）提取了 1216 个特征。采用最大相关性最小冗余（mRMR）和最小绝对收缩和选择算子（LASSO）算法构建 Radscore。随后评估并验证了诊断性能。比较了受试者工作特征（ROC）曲线和决策曲线分析（DCA），以评估联合模型的增量价值。
从 18 个特征中得到的 Radscore 在训练集中具有良好的鉴别能力，曲线下面积（AUC）为 0.834，而在验证集中为 0.770。联合模型在训练集中的 AUC 为 0.901（95%置信区间（95%CI）：0.865-0.938），在验证集中为 0.833（95%CI：0.788-0.878）。两者均优于 MSKCC 或影像学 Radscore 单独使用（P<0.05）。DCA 表明，联合模型在临床实用性方面优于其他模型。
术前超声的 Radscore 可提高 MSKCC 列线图对乳腺癌 SLN 转移预测的准确性。
