Min-Max Similarity: A Contrastive Semi-Supervised Deep Learning Network for Surgical Tools Segmentation

Segmentation of images is a popular topic in medical AI. This is mainly due to the difficulty to obtain a significant number of pixel-level annotated data to train a neural network. To address this issue, we proposed a semi-supervised segmentation network based on contrastive learning. In contrast to the previous state-of-the-art, we introduce Min-Max Similarity (MMS), a contrastive learning form of dual-view training by employing classifiers and projectors to build all-negative, and positive and negative feature pairs respectively to formulate the learning problem as solving min-max similarity problem. The all-negative pairs are used to supervise the networks learning from different views and make sure to capture general features, and the consistency of unlabeled predictions is measured by pixel-wise contrastive loss between positive and negative pairs. To quantitative and qualitative evaluate our proposed method, we test it on two public endoscopy surgical tool segmentation datasets and one cochlear implant surgery dataset which we manually annotate the cochlear implant in surgical videos. The segmentation performance (dice coefficients) indicates that our proposed method outperforms state-of-the-art semi-supervised and fully supervised segmentation algorithms consistently. And our semi-supervised segmentation algorithm can successfully recognize unknown surgical tools and provide good predictions. Also, our MMS could achieve about 40 frames per second (fps) and suitable to deal with the real-time video segmentation.