深度学习MMV-Net模型对乳腺X线良性和恶性肿块的分类效能
The classification performance of MMV-Net model for benign and malignant masses on X-ray mammography using deep learning
李家豪 1柏家和 1兰婕 1李海霞 2张岩 3孙江宏4
作者信息
- 1. 哈尔滨医科大学附属肿瘤医院(哈尔滨 150081)
- 2. 哈尔滨医科大学附属肿瘤医院超声科
- 3. 哈尔滨工业大学生命科学与技术学院
- 4. 哈尔滨医科大学附属肿瘤医院影像中心
- 折叠
摘要
目的 构建基于乳腺X线多视图的深度学习框架(Network based on mammography multiple views,MMV-Net),评价模型对乳腺良性和恶性肿块的分类效能.方法 回顾性分析 2018-2020 年哈尔滨医科大学附属肿瘤医院 1 585 例乳腺X线图像数据集,其中良性 806 例,恶性779 例,按8∶2 分为训练集(n=1268)和测试集(n=317),并按照5 折交叉验证对训练集进行分层,采用集成的DDSM数据集和INBreast数据集作为外部测试集(n=1 645)来评估模型性能.输入层每个病例包含 4 个视图,通过删除ResNet22 网络模型的最后两层网络结构并加入平均池化层作为特征提取层,以及分别加入全连接层和softmax激活函数作为决策层构建MMV-Net模型,使用贝叶斯超参数优化.比较MMV-Net、MFA-Net和集成Inception V4 模型在AUC值、准确率、精确率、召回率和F1 分数上的表现.结果 MMV-Net模型在测试集上区分良性和恶性肿块的AUC值为0.913,MFA-Net的AUC为0.882,Inception V4 的AUC为0.865;MMV-Net模型的准确率和精确率等评估指标也高于其他两种模型.结论 基于乳腺X线多视图的深度学习MMV-Net模型有助于乳腺良性和恶性肿块的分类.
Abstract
Objective The MMV-Net,a deep learning framework based on mammogram multiple views,was constructed to evaluate the classification performance of the model for benign and malignant masses.Methods A retrospective analysis was conduc-ted on a dataset of 1 585 breast X-ray images from Harbin Medical University Cancer Hospital from 2018 to 2020,including 806 be-nign cases and 779 malignant cases.The dataset was divided into the training set(n=1268)and the test set(n=317)according to an 8∶2 ratios,and the training set was stratified according to the 5-fold cross validation.The integrated DDSM dataset and INBreast dataset were used as external test sets(n=1645)to evaluate the model performance.Each case in the input layer contained 4 views.The MMV-Net model was constructed by removing the last two layers of the ResNet22 network structure and adding an average poo-ling layer as the feature extraction layer,as well as fully connection layer and softmax activation function as the decision layers.Bayes-ian hyperparameter optimization was used.The performance of MMV-Net,MFA Net,and ensemble inception V4 models in AUC val-ues,accuracy,precision,recall and F1 scores were compared.Results The AUC values of MMV-Net model for distinguishing benign and malignant masses on the test set were 0.913,0.882 for MFA-Net,and 0.865 for inception V4.The accuracy and precision evalu-ation metrics of the MMV-Net model were also higher than the other two models.Conclusion The deep learning MMV-Net model based on multiple views of mammogram is helpful for the classification of benign and malignant breast masses.
关键词
深度学习/MMV-Net模型/乳腺X线摄影/肿块/分类Key words
Deep learning/MMV-Net model/Mammography/Tumor/Classification引用本文复制引用
基金项目
大学生创新训练国家级一般项目资助(202310226066国家级)
大学生创新训练国家级一般项目资助(202310226023校级)
出版年
2024
NETL
NSTL
万方数据