Local contrastive loss with pseudo-label based self-training for semi-supervised medical image segmentation

Abstract

Supervised deep learning-based methods yield accurate results for medical image segmentation. However, they require large labeled datasets for this, and obtaining them is a laborious task that requires clinical expertise. Semi/selfsupervised learning-based approaches address this limitation by exploiting unlabeled data along with limited annotated data. Recent self-supervised learning methods use contrastive loss to learn good global level representations from unlabeled images and achieve high performance in classification tasks on popular natural image datasets like ImageNet. In pixel-level prediction tasks such as segmentation, it is crucial to also learn good local level representations along with global representations to achieve better accuracy. However, the impact of the existing local contrastive loss-based methods remains limited for learning good local representations because similar and dissimilar local regions are defined based on random augmentations and spatial proximity; not based on the semantic label of local regions due to lack of large-scale expert annotations in the semi/self-supervised setting. In this paper, we propose a local contrastive loss to learn good pixel level features useful for segmentation by exploiting semantic label information obtained from pseudo-labels of unlabeled images alongside limited annotated images. In particular, we define the proposed loss to encourage similar representations for the pixels that have the same pseudo-label/ label while being dissimilar to the representation of pixels with different pseudo-label/label in the dataset. We perform pseudo-label based self-training and train the network by jointly optimizing the proposed contrastive loss on both labeled and unlabeled sets and segmentation loss on only the limited labeled set. We evaluated the proposed approach on three public medical datasets of cardiac and prostate anatomies, and obtain high segmentation performance with a limited labeled set of one or two 3D volumes. Extensive comparisons with the state-of-the-art semi-supervised and data augmentation methods and concurrent contrastive learning methods demonstrate the substantial improvement achieved by the proposed method. The code is made publicly available at https://github.com/krishnabits001/pseudo_label_contrastive_training.

MaNi: Maximizing Mutual Information for Nuclei Cross-Domain Unsupervised Segmentation

“…Our work maximizes the MI between similar classes using labeled source images and pseudo-labeled target images. Concurrent to our work, Chaitanya et al [3] proposed an end-to-end semi-segmentation framework by defining a local pixel-level contrastive loss between pseudo-labels of unlabeled sets and limited labeled sets. They randomly sample pixels from each image to address the computational limitations of running CL for all the pixels.…”

“…The impact of MI loss dilutes 1 Other approach results reported using best self-implementation 2 Other approach results reported from [27] with reduction in weight, leading to a drop in performance. Further, we study how replacing the average pooling approach with the max-pooling approach or sampling random pixels approach (N=4), as done in [3], impacts performance. Average pooling performs superior to other approaches.…”

Rethinking Semi-Supervised Medical Image Segmentation: A Variance-Reduction Perspective

“…The past five years have seen tremendous progress related to CL in medical image segmentation [15,16,42,79,81,84,85,87,88,91], and it becomes increasingly important to improve representation in label-scarcity scenarios. The key idea in CL [21,39,41,59] is to learn representations from unlabeled data that obey similarity constraints by pulling augmented views of the same samples closer in a representation space, and pushing apart augmented views of different samples.…”