Masatoshi Abe

It is shown that self-supervised learning maintains high classification performance when labelled training data are limited and consistently outperforms ImageNet pre-training, which may facilitate the development of more efficient artificial intelligence-assisted cytology systems.