Segmenting unknown domains using a model trained in the source domain still faces challenges. Although some approaches tried to resolve the problem through various data generation and network architecture designs, they cannot achieve satisfactory segmentation results compared with single domain segmentation of consistent data distribution. Therefore, we propose a data augmentation method based on amplitude perturbation to expand the distribution of data types, thereby covering target data. A feature suppression strategy is proposed to reduce the network's over-reliance on important features of the source domain data to improve generalization performance. In addition, we design a luminance contrast consistency (LCC) learning module to harmonize the data styles between different domains and a multiscale convolutional attention (MSCA) module to enhance the network's perception of small target objects and improve the segmentation performance of the model, which further improves segmentation performance. Our method achieves the state-of-the-art (SOTA) results on two public datasets of ATLAS2.0 and Prostate. The code is available at https://github.com/butterflyGN/DGSFTAFS.

Semi-supervised learning for medical image segmentation presents a unique challenge of efficiently using limited labeled data while leveraging abundant unlabeled data. Despite advancements, existing methods often do not fully exploit the potential of the unlabeled data for enhancing model robustness and accuracy. In this paper, we introduce CrossMatch, a novel framework that integrates knowledge distillation with dual perturbation strategies, image-level and feature-level, to improve the model's learning from both labeled and unlabeled data. CrossMatch employs multiple encoders and decoders to generate diverse data streams, which undergo self-knowledge distillation to enhance the consistency and reliability of predictions across varied perturbations. Our method significantly surpasses other state-of-the-art techniques in standard benchmarks by effectively minimizing the gap between training on labeled and unlabeled data and improving edge accuracy and generalization in medical image segmentation. The efficacy of CrossMatch is demonstrated through extensive experimental validations, showing remarkable performance improvements without increasing computational costs.

In this paper, we focus on the empirical alignment challenge between labeled and unlabeled data in semi-supervised medical image segmentation. When labeled and unlabeled data are poorly aligned, the network struggles to fully leverage knowledge from the labeled data. To address this, we propose an efficient and streamlined approach called "DCA-Net," which integrates a frequency-domain data augmentation module Style Transfer Module (STM) and Bidirectional Copy-Paste (BCP) to effectively reduce the distribution gap between labeled and unlabeled data. Additionally, we combine knowledge distillation with semi-supervised learning to encourage deeper feature learning and more stable model behavior. Experiments with DCA-Net on the LA and ACDC datasets achieve state-of-the-art (SOTA) results.

With the development of medical image segmentation technology, high-quality automatic segmentation methods, particularly within semi-supervised learning frameworks, have become a research hotspot. This study introduces a new semi-supervised medical image segmentation algorithm called SymMatch. The algorithm effectively leverages limited labeled data along with a large amount of unlabeled data through a symmetrical network structure and knowledge distillation techniques. SymMatch applies a spectrum of perturbations, from weak to strong, at both image and feature levels, effectively leveraging the potential of unlabeled data. Additionally, by incorporating a bi-scale distillation loss, the model’s robustness and accuracy in handling complex medical imaging data are further enhanced. Experimental results show that SymMatch demonstrates superior performance across multiple recognized medical imaging datasets (such as ACDC, LA and PanNuke). Notably, even with very limited labeled data, it maintains high segmentation accuracy. These achievements not only advance the development of semi-supervised medical image segmentation technology but also provide new ideas and methods for future research in related technologies. Code is available at https://github.com/AiEson/SymMatch.

Semi-supervised medical image segmentation still faces challenges although it is able to obtain better segmentation results using a small amount of labeled data and a large amount of unlabeled data. Despite the progress made by current methods in utilizing unlabeled data, they fail to exploit the full potential of labeled data in terms of improving model performance. In this paper, we propose a semi-supervised segmentation method, DistillMatch, that incorporates knowledge distillation and feature perturbation, which efficiently transfers knowledge between labeled and unlabeled data, thus making full use of the information of labeled data to improve segmentation results. DistillMatch consists of several key components: the Self-Training process based on knowledge distillation and feature perturbation, the Deterministic Knowledge Transfer (DKT) strategy, and the introduction of Teacher Assistant (TA), which work together to optimize model performance. Extensive experiments on two benchmark datasets demonstrate that our method outperforms the current state-of-the-art (SOTA) approaches, especially in terms of edge accuracy and model generalization capabilities. We also show how this performance improvement can be achieved without sacrificing computational efficiency through an effective multi-decoder implementation strategy. These experimental results not only demonstrate the effectiveness of our approach, but also highlight its practical value in medical image segmentation tasks. Code is available at https://github.com/AiEson/DistillMatch.

Abstract Acute ischemic stroke is a common threat to human health and may obtain timely treatment by fast localizing and quantitatively evaluating the lesions. Most CNN‐based methods try to segment and measure the lesions, however, they require a training on a large number of labeled subjects that are labor‐intensive and time‐consuming to obtain. In this paper, a method is proposed that can combine limited labeled subjects with abundant unlabeled subjects to alleviate the problem. The proposed method consists of two stages: stepwise learning process and segmentation process. Stepwise learning is used to obtain the pretrained encoder. The pretrained encoder and the proposed decoder are connected into a new end‐to‐end segmentation network, which is retrained on the labeled subjects in the segmentation process. By using 5 labeled subjects and 79 unlabeled subjects, the proposed method achieves a mean dice coefficient of 0.6630.205, a mean average symmetric surface distance (ASSD) of 2.17 mm and a mean 95 percentile Hausdorff distance (HD) of 18.38 mm on a clinical MR dataset with 179 subjects. More importantly, it achieves lesion‐wise F 1 score of 0.857 and a subject‐wise detection rate of 0.966.