Department of Medical Ultrasound, Sun Yat-sen University Cancer Center, State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Guangzhou 510000, China (J.-X.H., X.-Y.W., Y.-F.X., M.-J.W., L.-Z.L., X.-Q.P.).
School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China (J.S., S.-S.D., H.-L.Z.).
Acad Radiol. 2023 Sep;30 Suppl 2:S50-S61. doi: 10.1016/j.acra.2023.03.036. Epub 2023 Jun 1.
To carry out radiomics analysis/deep convolutional neural network (CNN) based on B-mode ultrasound (BUS) and shear wave elastography (SWE) to predict response to neoadjuvant chemotherapy (NAC) in breast cancer patients.
In this prospective study, 255 breast cancer patients who received NAC between September 2016 and December 2021 were included. Radiomics models were designed using a support vector machine classifier based on US images obtained before treatment, including BUS and SWE. And CNN models also were developed using ResNet architecture. The final predictive model was developed by combining the dual-modal US and independently associated clinicopathologic characteristics. The predictive performances of the models were assessed with five-fold cross-validation.
Pretreatment SWE performed better than BUS in predicting the response to NAC for breast cancer for both the CNN and radiomics models (P < 0.001). The predictive results of the CNN models were significantly better than the radiomics models, with AUCs of 0.72 versus 0.69 for BUS and 0.80 versus 0.77 for SWE, respectively (P = 0.003). The CNN model based on the dual-modal US and molecular data exhibited outstanding performance in predicting NAC response, with an accuracy of 83.60% ± 2.63%, a sensitivity of 87.76% ± 6.44%, and a specificity of 77.45% ± 4.38%.
The pretreatment CNN model based on the dual-modal US and molecular data achieved excellent performance for predicting the response to chemotherapy in breast cancer. Therefore, this model has the potential to serve as a non-invasive objective biomarker to predict NAC response and aid clinicians with individual treatments.
利用 B 型超声（BUS）和剪切波弹性成像（SWE）的放射组学分析/深度卷积神经网络（CNN），预测乳腺癌患者新辅助化疗（NAC）的反应。
本前瞻性研究纳入了 2016 年 9 月至 2021 年 12 月期间接受 NAC 的 255 例乳腺癌患者。基于治疗前获得的 US 图像，设计了基于支持向量机分类器的放射组学模型，包括 BUS 和 SWE。并使用 ResNet 架构开发了 CNN 模型。最终的预测模型是通过结合双模态 US 和独立相关的临床病理特征来建立的。采用五重交叉验证评估模型的预测性能。
在 CNN 和放射组学模型中，预处理 SWE 在预测乳腺癌对 NAC 的反应方面均优于 BUS（P<0.001）。CNN 模型的预测结果明显优于放射组学模型，BUS 的 AUC 分别为 0.72 和 0.69，SWE 的 AUC 分别为 0.80 和 0.77（P=0.003）。基于双模态 US 和分子数据的 CNN 模型在预测 NAC 反应方面表现出色，准确率为 83.60%±2.63%，灵敏度为 87.76%±6.44%，特异性为 77.45%±4.38%。
基于双模态 US 和分子数据的预处理 CNN 模型在预测乳腺癌化疗反应方面表现出优异的性能。因此，该模型有可能成为一种非侵入性的客观生物标志物，用于预测 NAC 反应，辅助临床医生进行个体化治疗。
