Accelerating GluCEST imaging using deep learning for B₀ correction.

Area of Science:

• Neuroimaging
• Medical Physics
• Artificial Intelligence in Medicine

Background:

• Glutamate-weighted Chemical Exchange Saturation Transfer (GluCEST) MRI is a noninvasive brain imaging technique.
• Current GluCEST MRI is limited by long acquisition times and low signal-to-noise ratio (SNR) due to field inhomogeneity and small signal differences.

Purpose of the Study:

• To address the limitations of conventional GluCEST MRI.
• To develop a novel deep learning (DL)-based algorithm for faster and more accurate GluCEST imaging.

Main Methods:

• A DL algorithm utilizing wide activation neural network blocks was proposed.
• B0 correction involved reduced saturation offset acquisitions.
• Deep residual networks were trained to map sparse CEST-weighted images to a target spectrum.

Main Results:

• The DL-based methods outperformed traditional approaches visually and quantitatively.
• The wide activation blocks method achieved the highest Structural Similarity Index (SSIM) of 0.84 and peak SNR (PSNR) of 25dB.
• Significant SNR increases (over 8dB) were observed in regions of interest.

Conclusions:

• The novel DL method effectively reduces GluCEST imaging time by approximately 50%.
• The algorithm yields higher SNR compared to current state-of-the-art methods.
• This approach offers a more practical and sensitive solution for brain glutamate mapping.