基于纵向时间深度网络融合的乳腺癌新辅助化疗疗效预测  

Prediction of Breast Cancer Neoadjuvant Chemotherapy Based on Longitudinal Time Depth Network Fusion

作　　者：薛泰龙 范明 陈淑君[2] 厉力华[1] Xue Tailong;Fan Ming;Chen Shujun;Li Lihua(Institute of Biomedical Enginering and Instrument,Hangzhou Dianzi Unierity,Hangzhou 310018,China;Department of Rodiology,Zhejang Caner Hospital,Hangzhou 310022,China)

机构地区：[1]杭州电子科技大学生物医学工程与仪器研究所,杭州310018 [2]浙江省肿瘤医院放射科,杭州310022

出　　处：《中国生物医学工程学报》2022年第2期186-194,共9页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金(61871428);浙江省自然科学基金(J19H180004);浙江省公益技术应用社会发展项目(LGF18H180006)。

摘　　要：新辅助化疗提高了乳腺癌的治愈率,但并不是对所有患者都有效,准确预测化疗疗效可以为患者治疗方案的制定提供参考价值。本研究使用深度学习的方法,融合纵向时间的动态增强磁共振成像(DCE-MRI)的影像特征对新辅助化疗疗效进行预测。分析164例进行了乳腺癌新辅助化疗患者的DCE-MRI影像,从每例患者影像数据集中挑选肿瘤最大径及上下2张切片以扩充数据量至442例,并随机划分为训练集312例,测试集130例。DCE-MRI影像共6个序列,分割每个序列的乳房区域,去除皮肤和胸腔,使用深度学习模型分别根据化疗前影像、2个疗程化疗后影像、化疗前和2个疗程化疗后影像相融合对新辅助化疗疗效进行预测,并绘制预测结果的ROC曲线,计算对应曲线下面积(AUC)评估模型的分类性能。深度学习模型对化疗前影像、2个疗程化疗后影像的疗效预测的最佳AUC分别为0.775和0.808,融合化疗前和2个疗程化疗后影像对疗效进行预测的最佳AUC为0.863,预测效果优于仅使用化疗前的影像。实验结果表明,相较于单独使用化疗前影像,融合使用纵向时间的影像可以提高对新辅助化疗疗效的预测性能。Neoadjuvant chemotherapy can improve the cure rate of breast cancer, but it is not effective for all patients. Accurate prediction of chemotherapy efficacy can provide reference for physicians to formulate treatment protocols. This study used deep learning to integrate the image characteristics of longitudinal time dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI) to predict the efficacy of neoadjuvant chemotherapy. We analyzed 164 DCE-MRI images of patients who underwent neoadjuvant chemotherapy for breast cancer, and selected the maximum tumor diameter and two upper and lower slices from each patient′s image data set to expand the data to 442 cases that were randomly divided into 312 cases in the training set and 130 cases in the test set. DCE-MRI images had 6 sequences in total. Segmented the breast area of each sequence and removed the skin and chest cavity. Using deep learning model, the efficacy of neoadjuvant chemotherapy was predicted based on the images before chemotherapy, after 2 courses of chemotherapy and both of them, respectively. We drew the ROC curve of the prediction results and calculated the area under the curve(AUC) to evaluate the classification performance of the model. The best AUC of deep learning model for predicting the efficacy of the images before chemotherapy and the images after two courses of chemotherapy was 0.775 and 0.808 respectively, and the best AUC for predicting the efficacy of the fusion of images before chemotherapy and images after 2 courses of chemotherapy was 0.863, which was better than using the images before chemotherapy. The experimental results showed that compared with the existing approach of using the images before chemotherapy, using the fusion of longitudinal time images could improve the prediction performance of neoadjuvant chemotherapy.

关 键 词：乳腺癌 深度学习 新辅助化疗疗效 动态增强磁共振成像 纵向时间分析 

分 类 号：R318[医药卫生—生物医学工程]