论文标题
对比性质心监督减轻了医学图像分类中的域转移
Contrastive Centroid Supervision Alleviates Domain Shift in Medical Image Classification
论文作者
论文摘要
基于深度学习的医学成像分类模型通常会遭受域转移问题的困扰，在训练数据和现实世界数据的成像设备制造商，图像获取协议，患者人群等方面的分类性能会下降。我们提出了特征质心对比度学习（FCCL），这可以通过在实例和类别中进行对比型损失在对比的培训期间通过额外的监督性能提高目标域分类性能。与当前的无监督域的适应和域的概括方法相比，FCCL的性能更好，而仅需要来自单个源域而不是目标域的标记图像数据。我们通过广泛的实验来验证FCCL可以在至少三个成像方式上实现卓越的性能，即眼底照片，皮肤镜图像和H＆E组织图像。
Deep learning based medical imaging classification models usually suffer from the domain shift problem, where the classification performance drops when training data and real-world data differ in imaging equipment manufacturer, image acquisition protocol, patient populations, etc. We propose Feature Centroid Contrast Learning (FCCL), which can improve target domain classification performance by extra supervision during training with contrastive loss between instance and class centroid. Compared with current unsupervised domain adaptation and domain generalization methods, FCCL performs better while only requires labeled image data from a single source domain and no target domain. We verify through extensive experiments that FCCL can achieve superior performance on at least three imaging modalities, i.e. fundus photographs, dermatoscopic images, and H & E tissue images.
