Liu Liangsheng, He Shanshan, Niu Zhenhua, Yin Rui, Guo Yijun, Dou Zhaoxiang, Ma Wenjuan, Ye Zhaoxiang, Lu Hong
Department of Breast Imaging, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Key Laboratory of Breast Cancer Prevention and Therapy, Tianjin Medical University, Ministry of Education, Key Laboratory of Cancer Prevention and Therapy, Tianjin's Clinical Research Center for Cancer, Tianjin, China.
Department of Breast Reconstruction, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Key Laboratory of Breast Cancer Prevention and Therapy, Tianjin Medical University, Ministry of Education, Key Laboratory of Cancer Prevention and Therapy, Tianjin's Clinical Research Center for Cancer, Tianjin, China.
Gland Surg. 2024 May 30;13(5):640-653. doi: 10.21037/gs-23-509. Epub 2024 May 27.
Breast-conserving surgery (BCS) stands as the favored modality for treating early-stage breast cancer. Accurately forecasting the feasibility of BCS preoperatively can aid in surgical planning and reduce the rate of switching of surgical methods and reoperation. The objective of this study is to identify the radiomics features and preoperative breast magnetic resonance imaging (MRI) characteristics that are linked with positive margins following BCS in patients with breast cancer, with the ultimate aim of creating a predictive model for the feasibility of BCS.
This study included a cohort of 221 pretreatment MRI images obtained from patients with breast cancer. A total of seven MRI semantic features and 1,561 radiomics features of lesions were extracted. The feature subset was determined by eliminating redundancy and correlation based on the features of the training set. The least absolute shrinkage and selection operator (LASSO) logistic regression was then trained with this subset to classify the final BCS positive and negative margins and subsequently validated using the test set.
Seven features were significant in the discrimination of cases achieving positive and negative margins. The radiomics signature achieved area under the curve (AUC), accuracy, sensitivity, and specificity of 0.760 [95% confidence interval (CI): 0.630, 0.891], 0.712 (95% CI: 0.569, 0.829), 0.882 (95% CI: 0.623, 0.979) and 0.629 (95% CI: 0.449, 0.780) in the test set, respectively. The combined model of radiomics signature and background parenchymal enhancement (BPE) demonstrated an AUC, accuracy, sensitivity, and specificity of 0.759 (95% CI: 0.628, 0.890), 0.654 (95% CI: 0.509, 0.780), 0.679 (95% CI: 0.476, 0.834) and 0.625 (95% CI: 0.408, 0.804).
The combination of preoperative MRI radiomics features can well predict the success of breast conserving surgery.
保乳手术（BCS）是治疗早期乳腺癌的首选方式。术前准确预测BCS的可行性有助于手术规划，并降低手术方式转换和再次手术的发生率。本研究的目的是识别与乳腺癌患者BCS术后切缘阳性相关的影像组学特征和术前乳腺磁共振成像（MRI）特征，最终建立一个预测BCS可行性的模型。
本研究纳入了221例乳腺癌患者治疗前的MRI图像队列。共提取了7个MRI语义特征和1561个病变的影像组学特征。基于训练集的特征消除冗余和相关性来确定特征子集。然后使用该子集训练最小绝对收缩和选择算子（LASSO）逻辑回归，以对最终BCS的切缘阳性和阴性进行分类，并随后使用测试集进行验证。
7个特征在区分切缘阳性和阴性病例中具有显著性。影像组学特征在测试集中的曲线下面积（AUC）、准确性、敏感性和特异性分别为0.760 [95%置信区间（CI）：0.630，0.891]、0.712（95%CI：0.569，0.829）、0.882（95%CI：0.623，0.979）和0.629（95%CI：0.449，0.780）。影像组学特征与背景实质强化（BPE）的联合模型的AUC、准确性、敏感性和特异性分别为0.759（95%CI：0.628，0.890）、0.654（95%CI：0.509，0.780）、0.679（95%CI：0.476，0.834）和0.625（95%CI：0.408，0.804）。
术前MRI影像组学特征的组合能够很好地预测保乳手术的成功。
