Voxel-wise Adversarial Semi-supervised Learning for Medical Image Segmentation

TL;DR

This paper presents a novel voxel-wise adversarial semi-supervised learning method for medical image segmentation that effectively captures local and global features, improving performance over existing approaches.

Contribution

Introduces a voxel-wise adversarial learning approach that embeds multi-layer features and models class-specific context relations, enhancing semi-supervised segmentation accuracy.

Findings

Outperforms state-of-the-art semi-supervised methods on multi-organ datasets.

Enables well-distributed and separated feature space for labeled and unlabeled data.

Improves representation stability and reduces information loss.

Abstract

Semi-supervised learning for medical image segmentation is an important area of research for alleviating the huge cost associated with the construction of reliable large-scale annotations in the medical domain. Recent semi-supervised approaches have demonstrated promising results by employing consistency regularization, pseudo-labeling techniques, and adversarial learning. These methods primarily attempt to learn the distribution of labeled and unlabeled data by enforcing consistency in the predictions or embedding context. However, previous approaches have focused only on local discrepancy minimization or context relations across single classes. In this paper, we introduce a novel adversarial learning-based semi-supervised segmentation method that effectively embeds both local and global features from multiple hidden layers and learns context relations between multiple classes. Our voxel-wise adversarial learning method utilizes a voxel-wise feature discriminator, which considers multilayer voxel-wise features (involving both local and global features) as an input by embedding class-specific voxel-wise feature distribution. Furthermore, we improve our previous representation learning method by overcoming information loss and learning stability problems, which enables rich representations of labeled data. Our method outperforms current best-performing state-of-the-art semi-supervised learning approaches on the image segmentation of the left atrium (single class) and multiorgan datasets (multiclass). Moreover, our visual interpretation of the feature space demonstrates that our proposed method enables a well-distributed and separated feature space from both labeled and unlabeled data, which improves the overall prediction results.