SMDFnet: Saliency multiscale dense fusion network for MRI and CT image fusion

Area of Science:

• Medical Imaging
• Computer Vision
• Image Processing

Background:

• Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) fusion enhances diagnostic accuracy but faces challenges.
• Artifacts and information loss in current MRI-CT fusion methods limit their effectiveness.
• Existing fusion strategies are complex and require further optimization.

Purpose of the Study:

• To develop an advanced MRI-CT image fusion technique.
• To address limitations of current fusion methods, including artifacts and information loss.
• To improve the quality and accuracy of fused medical images.

Main Methods:

• A saliency multi-scale dense fusion network is proposed for MRI-CT image fusion.
• Pre-trained networks extract depth information, mitigating noise and artifacts.
• A multi-scale dense network enhances features, followed by fusion using multiple loss functions.

Main Results:

• The proposed method significantly outperforms reference methods in objective indicators.
• Enhanced fusion results retain more salient information compared to existing techniques.
• The approach effectively overcomes noise and artifacts present in individual MRI and CT images.

Conclusions:

• The developed saliency multi-scale dense fusion network offers superior performance for MRI-CT image fusion.
• This method improves image quality and information preservation in medical diagnostics.
• The findings suggest a promising advancement for medical diagnosis and surgical planning.