Joint Image Reconstruction and Super-Resolution for Accelerated Magnetic Resonance Imaging

Area of Science:

• Medical Imaging
• Image Processing
• Biomedical Engineering

Background:

• Magnetic resonance (MR) imaging acceleration is crucial for reducing scan times.
• Undersampled and low-resolution k-space data acquisition are common strategies for acceleration.
• Current methods often address image reconstruction and super-resolution separately, leading to potential errors.

Purpose of the Study:

• To develop a novel framework for joint image reconstruction and super-resolution in MR imaging.
• To enable faster MR imaging while maintaining high image quality.
• To overcome limitations of sequential or separate processing of reconstruction and super-resolution.

Main Methods:

• A multi-task learning framework integrating a model-based reconstruction module and a super-resolution module was designed.
• The reconstruction module ensures data fidelity using acquired k-space data.
• A deep spatial feature transform was employed to enhance information transition between modules.

Main Results:

• The proposed joint framework demonstrated superior quantitative and qualitative performance on two datasets.
• Effective integration of image reconstruction and super-resolution was achieved.
• The method facilitates efficient image recovery and enables faster MR imaging.

Conclusions:

• The novel framework for joint MR image reconstruction and super-resolution offers significant advantages over existing methods.
• This approach enhances image quality and accelerates MR imaging acquisition.
• The integrated multi-task learning strategy with deep spatial feature transform is effective for MR image recovery.