Research progress of medical image fusion methods based on deep learning

VernacularTitle:基于深度学习的医学图像融合方法的研究进展
Author: Yi ZHANG ¹ ; Liu LIU ¹ ; Meng WANG ¹ ; Wenhui XIE ¹
Author Information

1. 上海交通大学医学院附属胸科医院核医学科，上海　200030
Publication Type:Journal Article
Keywords: Deep learning; Image processing, computer-assisted; Trends
From: Chinese Journal of Nuclear Medicine and Molecular Imaging 2025;45(10):637-640
CountryChina
Language:Chinese
Abstract: Medical image fusion technology combines the advantages of different modal medical images to provide more comprehensive and precise imaging information, widely applied in clinical diagnosis and disease research. Traditional fusion methods rely on signal processing techniques, but they have significant limitations, especially in multi-modal image fusion, where extracting deep semantic information is challenging. With the introduction of deep learning, image fusion effects have been significantly enhanced through technologies such as convolutional neural networks (CNN), generative adversarial networks (GAN), deep feature extractors, and self-attention mechanisms. However, deep learning methods still face challenges such as data scarcity, modal heterogeneity, high computational resource requirements, mode collapse, and training instability. This paper analyzes these issues and explores corresponding solutions as well as potential future development directions.