Multi-Modal Iterative Refinement Network with K-space Posterior Correction for MRI reconstruction

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) is vital for clinical diagnosis but faces challenges with long acquisition times.
• Undersampling accelerates MRI but causes aliasing artifacts and loss of image detail.
• Reference-based reconstruction methods use multi-modal data but are limited by spatial misalignment and ignore k-space priors.

Purpose of the Study:

• To develop an advanced MRI reconstruction method addressing spatial misalignment and k-space limitations.
• To enhance accelerated MRI reconstruction quality using multi-modal anatomical information.
• To introduce a novel network for improved Magnetic Resonance Imaging reconstruction.

Main Methods:

• Proposed the Multi-Modal Iterative Refinement Network with K-space Posterior Correction (MMIR-Net).
• Developed an Iterative Refinement Network (IR-Net) with a Residual Registration Module (RRM) for progressive alignment.
• Incorporated a K-space Posterior Correction Module (KPCM) for physical prior correction and an Adaptive Fusion Module (AFM) for domain integration.

Main Results:

• MMIR-Net demonstrated superior performance compared to existing methods on IXI and fastMRI datasets.
• The method effectively handles various undersampling patterns and ratios.
• Achieved enhanced reconstruction quality in accelerated MRI.

Conclusions:

• The proposed MMIR-Net offers a novel and effective solution for multi-modal MRI reconstruction challenges.
• MMIR-Net successfully integrates image and k-space information for improved MRI.
• This work advances accelerated MRI reconstruction by addressing key limitations of current techniques.