Accelerating quantitative MR imaging with the incorporation of B₁ compensation using deep learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Quantitative MRI

Background:

• Quantitative MRI (qMRI) is crucial for data-driven medicine but limited by long acquisition times.
• Repetitive imaging and field map measurements contribute to lengthy qMRI protocols.
• Accelerating qMRI acquisition is essential for broader clinical and research applications.

Purpose of the Study:

• To develop and validate a deep learning strategy for accelerating quantitative T1 mapping in MRI.
• To enable accurate T1 map generation from highly undersampled, variable-contrast images with automatic B1 field correction.
• To reduce overall MRI acquisition time without compromising image fidelity.

Main Methods:

• A multi-step deep learning framework was employed, utilizing convolutional neural networks.
• Variable-contrast images were jointly reconstructed from undersampled data.
• T1 and B1 maps were predicted from reconstructed images using AI.
• The method was validated for T1 mapping of cartilage using extensive datasets.

Main Results:

• The deep learning strategy achieved high acceleration factors in quantitative MRI acquisition.
• Image fidelity was well-maintained despite significant undersampling and reduced contrast images.
• The method successfully compensated for radiofrequency field inhomogeneity.
• Validation demonstrated the effectiveness of the AI approach in cartilage T1 mapping.

Conclusions:

• The proposed deep learning method significantly accelerates quantitative MRI acquisition, particularly for T1 mapping.
• This AI-driven approach overcomes limitations of long scan times and B1 map measurements.
• The strategy shows broad applicability for quantifying other tissue properties like T2 and T1ρ.