Related Experiment Video
Updated: Oct 11, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Generative adversarial networks in medical image segmentation: A review
Siyi Xun1, Dengwang Li1, Hui Zhu2
1Shandong Key Laboratory of Medical Physics and Image Processing, Shandong Institute of Industrial Technology for Health Sciences and Precision Medicine, School of Physics and Electronics, Shandong Normal University, Jinan, Shandong, 250358, China.
Computers in Biology and Medicine
|December 5, 2021
Summary
Generative Adversarial Networks (GANs) enhance medical image segmentation accuracy. This review covers GAN origins, principles, variants, and applications in segmentation, highlighting future research needs.
Area of Science:
- Deep Learning
- Medical Imaging
- Computer Vision
Background:
- Generative Adversarial Network (GAN) introduced in 2014.
- GANs have gained significant academic and industrial attention.
- GANs improve medical image segmentation accuracy due to their generative capabilities.
Purpose of the Study:
- Review the origin, working principle, and variants of GANs.
- Summarize the latest developments in GAN-based medical image segmentation.
- Provide insights into the advantages, challenges, and future directions of GANs in this field.
Main Methods:
- Searched Google Scholar and PubMed for "segmentation", "medical image", and "GAN".
- Conducted additional searches on Semantic Scholar, Springer, arXiv, and top computer science conferences.
- Reviewed over 120 GAN-based medical image segmentation papers published before September 2021.
Main Results:
- Categorized and summarized papers by segmentation regions, imaging modality, and classification methods.
- Discussed the advantages and challenges of using GANs for medical image segmentation.
- Identified over 120 GAN-based architectures for medical image segmentation.
Conclusions:
- GANs and their variants have significantly improved medical image segmentation accuracy.
- Future research should focus on overcoming GAN instability, low repeatability, and uninterpretability.
- Gaining clinical and patient acceptance is crucial for GAN adoption in medical imaging.
