基于乳腺超声视频流和自监督对比学习的肿瘤良恶性分类系统

唐蕴芯, 廖梅, 张艳玲, 张建, 陈皓, 王炜

Breast tumor classification based on video stream and self⁃supervised contrastive learning

Yunxin Tang, Mei Liao, Yanling Zhang, Jian Zhang, Hao Chen, Wei Wang

表3 三胞胎网络和其他SOTA预训练模型的实验结果对比

Table 3 Experimental results of our Triplet Network and other SOTA models

Model	AUC	Sensitivity	Specificity
DenseNet169⁃三胞胎网络 (Hard Triplet Loss)	0.952	0.890	0.890
MoCo⁃v2	0.756	0.674	0.674
BYOL	0.764	0.676	0.676
SwAV	0.752	0.665	0.665
DenseNet161⁃ImageNet 预训练模型	0.899	0.820	0.820