Machine learning in Magnetic Resonance Imaging: Image reconstruction.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) is crucial for disease diagnosis and monitoring but is inherently slow.
• Existing acceleration techniques like compressed sensing face challenges with reconstruction time and image quality.
• Limited clinical adoption of accelerated MRI hinders its widespread application.

Purpose of the Study:

• To review current machine learning (ML) approaches for MRI image reconstruction.
• To discuss the drawbacks, clinical applications, and emerging trends of ML in MRI.
• To highlight how ML can address limitations in accelerated MRI acquisition.

Main Methods:

• Review of various ML-based reconstruction methods applied in k-space and/or image-space.
• Analysis of algorithms addressing challenges in compressed sensing MRI.
• Summary of ML techniques enabling faster computation and improved image quality.

Main Results:

• ML approaches demonstrate promising results in MRI reconstruction.
• AI enables the generation of natural-looking images with rapid computation.
• Significant improvements in MRI acquisition speed and diagnostic utility are achievable.

Conclusions:

• Machine learning offers a powerful solution to accelerate MRI acquisition and enhance image quality.
• ML-driven reconstruction is poised to overcome previous limitations of accelerated MRI.
• Future trends indicate wider clinical integration of AI in MRI workflows.