Dynamic MRI reconstruction with end-to-end motion-guided network.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Deep Learning

Background:

• Temporal correlation in dynamic MRI is crucial for understanding body motion, but current deep learning methods often ignore this information.
• Traditional motion-guided MRI reconstruction methods suffer from complex parameter tuning and lengthy processing times.
• Artifacts and blurring in dynamic MRI sequences hinder accurate motion analysis.

Purpose of the Study:

• To develop a novel deep learning approach for dynamic MRI reconstruction that incorporates motion information.
• To enhance image quality and reduce artifacts in dynamic MRI sequences.
• To overcome limitations of existing deep learning and traditional motion-guided reconstruction methods.

Main Methods:

• Proposed MODRN (Motion-guided Dynamic Reconstruction Network) and its end-to-end version MODRN(e2e).
• Decomposed the motion-guided optimization into Dynamic Reconstruction Network, Motion Estimation, and Motion Compensation modules.
• Integrated motion information directly into the deep neural network-based reconstruction process.

Main Results:

• MODRN and MODRN(e2e) significantly improved dynamic MRI reconstruction quality.
• The proposed methods demonstrated superior performance compared to state-of-the-art approaches in extensive experiments.
• Effective incorporation of motion information led to temporally coherent image sequences with reduced artifacts and blurring.

Conclusions:

• MODRN offers an effective deep learning framework for dynamic MRI reconstruction by leveraging motion information.
• The proposed approach provides a more efficient and accurate alternative to existing methods for dynamic MRI.
• Future work can explore further optimization and application of MODRN in various dynamic MRI scenarios.