Rapid MR relaxometry using deep learning: An overview of current techniques and emerging trends

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Quantitative Imaging
• Medical Physics

Background:

• MR relaxometry (T1, T2, T1ρ mapping) offers superior sensitivity to pathologies compared to conventional MRI.
• Deep learning (DL) is revolutionizing MRI research, yet its application in MR relaxometry is underexplored.
• Quantitative MRI provides tissue-specific information crucial for diagnosis and prognosis.

Purpose of the Study:

• To review deep learning applications for accelerating MR relaxometry.
• To discuss emerging DL techniques for enhancing MR relaxometry speed, image quality, and quantification robustness.
• To bridge the gap between conventional MR relaxometry acceleration methods and DL-based approaches.

Main Methods:

• Review of existing "classical" MR relaxometry acceleration techniques (spatiotemporal acceleration, model-based reconstruction, efficient parameter generation).
• Exploration of deep learning methodologies applicable to MR relaxometry.
• Analysis of how DL integrates with and improves upon conventional acceleration methods.

Main Results:

• Deep learning demonstrates significant potential to improve MR relaxometry efficiency and accuracy.
• DL techniques can address limitations of traditional acceleration methods.
• Emerging DL approaches offer pathways to faster and more robust quantitative MRI.

Conclusions:

• Deep learning holds immense promise for rapid and reliable MR relaxometry.
• Addressing current challenges in DL for MR relaxometry is crucial for clinical translation.
• Future research should focus on robust DL solutions for quantitative MRI.