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1585例乳腺X线图像数据集,其中良性...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1585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=1645)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。展开更多
The distribution characteristics of air voids in ultrathin asphalt friction course(UAFC) samples with different gradations and compaction methods were statistically analyzed using X-ray computed tomography(CT) and ima...The distribution characteristics of air voids in ultrathin asphalt friction course(UAFC) samples with different gradations and compaction methods were statistically analyzed using X-ray computed tomography(CT) and image analysis techniques. Based on the results, compared with the AC-5 sample, the OGFC-5mixture has a higher air void ratio, a larger air void size and a greater number of air voids, with the distribution of internal air voids being more uniform and their shapes being more rounded. The two-parameter Weibull function was applied to fit the gradation of air voids. The fitting results is good, and the function parameters are sensitive to changes in both mineral gradation and compaction method. Moreover, two homogeneity indices were proposed to evaluate the compaction uniformity of UAFC samples. Compared with the Marshall method,the SGC method is more conducive to improve the compaction uniformity of UAFC samples. The compaction method significantly influences the air void distribution characteristics and compaction uniformity of AC-5sample, but has a less significant impact on OGFC-5 sample. The experimental results in the study provides a solid foundation for further explorations on the internal structure and mixture design of UAFC.展开更多
文摘目的构建基于乳腺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1585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=1645)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。
基金Funded by Technology Innovation Demonstration Project of the Transportation Department of Yunnan Province (Science and Technology Education Section of Transport Department of Yunnan Province [2019](No.14)。
文摘The distribution characteristics of air voids in ultrathin asphalt friction course(UAFC) samples with different gradations and compaction methods were statistically analyzed using X-ray computed tomography(CT) and image analysis techniques. Based on the results, compared with the AC-5 sample, the OGFC-5mixture has a higher air void ratio, a larger air void size and a greater number of air voids, with the distribution of internal air voids being more uniform and their shapes being more rounded. The two-parameter Weibull function was applied to fit the gradation of air voids. The fitting results is good, and the function parameters are sensitive to changes in both mineral gradation and compaction method. Moreover, two homogeneity indices were proposed to evaluate the compaction uniformity of UAFC samples. Compared with the Marshall method,the SGC method is more conducive to improve the compaction uniformity of UAFC samples. The compaction method significantly influences the air void distribution characteristics and compaction uniformity of AC-5sample, but has a less significant impact on OGFC-5 sample. The experimental results in the study provides a solid foundation for further explorations on the internal structure and mixture design of UAFC.
