Highly-accelerated CEST MRI using frequency-offset-dependent k-space sampling and deep-learning reconstruction.

Area of Science:

• Magnetic Resonance Imaging
• Medical Physics
• Biomedical Engineering

Background:

• Chemical Exchange Saturation Transfer (CEST) MRI is a valuable technique for metabolic imaging.
• Accelerating CEST acquisition is crucial for clinical applications, especially in neuro-oncology.
• Current CEST methods face limitations in speed and reconstruction accuracy.

Purpose of the Study:

• To develop a highly accelerated CEST Z-spectral acquisition method.
• To implement a novel k-space sampling pattern and a deep learning-based reconstruction.
• To evaluate the performance of the proposed method in human brain datasets.

Main Methods:

• A customized k-space sampling pattern with frequency-offset-dependent (FOD) probability was designed.
• A convolution neural network (CNN) enhanced with a Partially Separable (PS) function (PSN) was used for reconstruction.
• Retrospective and prospective experiments were conducted on human brain datasets, including healthy subjects and brain tumor patients.

Main Results:

• The FOD sampling combined with PSN (FOD+PSN) significantly outperformed conventional methods in quantitative metrics (nMSE, SSIM).
• FOD+PSN consistently improved four contrast maps (MTRasym, MTRrex, amide, NOE) across acceleration factors (4-14).
• Prospective results demonstrated the feasibility of the accelerated CEST MRI for brain tumor patients.

Conclusions:

• The combination of FOD sampling and PSN reconstruction enables highly accelerated CEST MRI acquisition.
• This advanced technique shows potential for facilitating CEST metabolic MRI in brain tumor patients.
• The developed method offers a promising approach for faster and more accurate metabolic imaging.