MC-RED: A deep learning network for motion correction in 3D CEST imaging.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Chemical Exchange Saturation Transfer (CEST) imaging offers high sensitivity for molecular analysis.
• Patient motion is a significant challenge, compromising the reliability of quantitative CEST imaging.
• Developing robust motion correction techniques is crucial for advancing CEST applications.

Purpose of the Study:

• To develop and validate a deep learning-based motion correction method for 3D CEST imaging.
• To enhance image quality and improve the reliability of quantitative molecular analysis in CEST.
• To address the limitations imposed by patient motion in CEST imaging.

Main Methods:

• Introduction of MC-RED, a motion correction approach utilizing a residual encoding-decoding network.
• Incorporation of frequency-specific information using a 2D Gaussian distribution with static reference images.
• Generation of motion-free reference frames for registration and correction of motion-corrupted CEST images, validated on simulated and clinical data.

Main Results:

• MC-RED significantly reduces motion artifacts, especially in low-contrast regions near the water resonance.
• Enhanced image quality demonstrated by improved signal fidelity and spatial alignment.
• More accurate quantitative maps achieved through effective motion correction.

Conclusions:

• The deep learning-based MC-RED method effectively corrects motion artifacts in 3D CEST imaging.
• This approach holds significant potential for increasing the reliability of quantitative CEST analysis.
• MC-RED represents a valuable advancement for motion-sensitive quantitative imaging techniques.