Consistency and adversarial semi-supervised learning for medical image segmentation

Tang, Yongqiang; Wang, Shilei; Qu, Yuxun; Cui, Zhihua; Zhang, Wensheng

doi:10.1016/j.compbiomed.2023.107018

Cited by 7 publications

(1 citation statement)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The effectiveness of deep learning methods often hinges on the availability of large, accurately labeled datasets, which can be challenging to acquire in the medical image analysis domain. To address the high costs and time associated with obtaining detailed annotations like pixel-level segmentation masks, research has shifted towards Semi-Supervised Learning (SSL) [4,16,33,10,31] and Weakly-Supervised Learning (WSL) [15,35,20,13,34]. SSL focuses on training networks with a small set of pixel-level labeled data, whereas WSL employs simpler forms of annotations such as bounding boxes, checkmarks, and points to provide a feasible approach for training segmentation networks under limited-signal supervision.…”

Section: Introductionmentioning

confidence: 99%

SMESwin Unet: Merging CNN and Transformer for Medical Image Segmentation

Wang

Min

Shi

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Medical image segmentation is increasingly reliant on deep learning techniques, yet the promising performance often come with high annotation costs. This paper introduces Weak-Mamba-UNet, an innovative weakly-supervised learning (WSL) framework that leverages the capabilities of Convolutional Neural Network (CNN), Vision Transformer (ViT), and the cutting-edge Visual Mamba (VMamba) architecture for medical image segmentation, especially when dealing with scribble-based annotations. The proposed WSL strategy incorporates three distinct architecture but same symmetrical encoder-decoder networks: a CNNbased UNet for detailed local feature extraction, a Swin Transformerbased SwinUNet for comprehensive global context understanding, and a VMamba-based Mamba-UNet for efficient long-range dependency modeling. The key concept of this framework is a collaborative and crosssupervisory mechanism that employs pseudo labels to facilitate iterative learning and refinement across the networks. The effectiveness of Weak-Mamba-UNet is validated on a publicly available MRI cardiac segmentation dataset with processed scribble annotations, where it surpasses the performance of a similar WSL framework utilizing only UNet or Swin-UNet. This highlights its potential in scenarios with sparse or imprecise annotations. The source code is made publicly accessible.

show abstract