CAMP-Net: Consistency-Aware Multi-Prior Network for Accelerated MRI Reconstruction.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computer Vision

Background:

• Accelerated magnetic resonance imaging (MRI) reduces scan times by undersampling k-space data.
• Reconstructing high-quality images from highly undersampled data, especially preserving fine details, remains a significant challenge.
• Existing methods struggle with artifact removal and detail preservation in severely undersampled MRI.

Purpose of the Study:

• To propose and evaluate CAMP-Net, a novel unrolling-based network for accelerated MRI reconstruction.
• To address the challenge of restoring high-frequency image details in highly undersampled k-space data.
• To improve the overall reconstruction quality and mapping estimation in accelerated MRI.

Main Methods:

• CAMP-Net, a Consistency-Aware Multi-Prior Network, utilizes an unrolling-based approach.
• It integrates complementary multi-prior knowledge from image, k-space, and calibration domains across multiple slices.
• Key components include interleaved modules for image enhancement, k-space restoration, calibration consistency, and a frequency fusion module, incorporating a surface data fidelity layer.

Main Results:

• CAMP-Net demonstrates superior performance compared to state-of-the-art methods in accelerated MRI reconstruction.
• The method excels in preserving high-frequency details and reducing artifacts, particularly at high acceleration factors.
• Experimental results on public datasets show improved reconstruction quality and mapping estimation accuracy.

Conclusions:

• CAMP-Net effectively leverages multi-domain priors and multi-slice information for robust accelerated MRI reconstruction.
• The proposed network offers a significant advancement in addressing the trade-off between artifact removal and fine detail preservation.
• CAMP-Net shows strong generalizability and robustness across different acceleration factors and sampling patterns.