Scan-Specific Unsupervised Highly Accelerated Non-Cartesian CEST Imaging Using Implicit Neural Representation and...

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Biomedical Engineering
• Artificial Intelligence in Medical Imaging

Background:

• Chemical exchange saturation transfer (CEST) is a valuable MRI technique.
• Current CEST imaging suffers from long scan times, limiting clinical applicability.
• Accelerating CEST acquisition is crucial for widespread adoption.

Purpose of the Study:

• To develop a novel deep learning algorithm for accelerating CEST imaging.
• To improve the efficiency and image quality of steady-state pulsed CEST acquisition.
• To validate the proposed method against existing state-of-the-art algorithms.

Main Methods:

• A scan-specific unsupervised deep learning algorithm using hybrid-feature hash encoding implicit neural representation.
• Incorporation of low-rank and weighted joint sparsity priors in spatial and Z-spectral domains.
• Application to steady-state pulsed CEST imaging with a golden-angle stack-of-stars trajectory.

Main Results:

• The proposed method, INRESP, outperformed other leading algorithms in retrospective acceleration experiments on human brain data.
• Prospective acceleration experiments demonstrated results comparable to fully-sampled images.
• Significant reduction in error and enhancement of image quality were observed at high acceleration factors.

Conclusions:

• The INRESP algorithm efficiently accelerates CEST MRI scans.
• It offers superior performance and image quality compared to existing methods, especially at high acceleration rates.
• Its training database-free nature makes it a promising tool for diverse CEST imaging applications.