Contrastive Learning Guided Fusion Network for Brain CT and MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Medical image fusion enhances diagnostic accuracy by integrating information from multiple imaging modalities.
• Current fusion techniques often require complex training or lack efficiency.

Purpose of the Study:

• To develop an efficient and accurate unsupervised medical image fusion network using contrastive learning.
• To improve the diagnostic information provided by fused CT and MRI images.

Main Methods:

• Introduced CLGFusion, a novel contrastive learning-guided network for CT and MRI fusion.
• Employed dual encoding branches with inter-branch interaction and an exponential moving average strategy.
• Integrated contrastive learning without negative samples, utilizing feature differences and structural similarity loss.

Main Results:

• CLGFusion demonstrated comparable performance to state-of-the-art methods.
• The unsupervised, end-to-end model achieved accurate and efficient image fusion.
• Experimental validation confirmed the effectiveness of the proposed fusion approach.

Conclusions:

• CLGFusion offers a promising unsupervised approach for medical image fusion.
• The contrastive learning strategy effectively guides the fusion process for improved diagnostic utility.
• The method enhances the precision and detail of fused medical images.