Hierarchical Multi-Scale Feature Fusion Network with Implicit Neural Representation and Mamba for Cross-Modality MRI...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) provides multimodal images essential for medical analysis.
• Simultaneous acquisition of all MRI modalities is often impractical due to patient discomfort, time, and cost.
• Synthesizing missing MRI modalities from available ones offers an effective solution.

Purpose of the Study:

• To propose HMF-MambaINR, a novel network for cross-modality MRI synthesis.
• To address the challenge of missing MRI modalities in clinical settings.
• To enhance medical image analysis through accurate MRI modality generation.

Main Methods:

• Developed HMF-MambaINR, a hierarchical multi-scale feature fusion network.
• Integrated Mamba-based Selective State Space Modeling (SSM) for long-range dependencies.
• Utilized implicit neural representation (INR) for continuous spatial reconstruction.
• Employed Multi-Feature Extraction Block (MFEB) and Modulation Fusion Module (MFM) for feature fusion.

Main Results:

• HMF-MambaINR outperformed state-of-the-art CNN-, Transformer-, and Mamba-based methods in synthesizing missing MRI modalities.
• Synthesized MRI images demonstrated high quality, accurate contrast, and precise structural contours.
• Radiologists provided positive feedback on the clinical utility of the generated images.

Conclusions:

• HMF-MambaINR presents a powerful approach for cross-modality MRI synthesis.
• The method shows significant potential for practical clinical applications in medical image analysis.
• Accurate synthesis of missing MRI modalities can improve diagnostic capabilities and reduce scanning burdens.