Retrospective Motion Correction in Multishot MRI using Generative Adversarial Network.

Area of Science:

• Medical Imaging
• Artificial Intelligence

Background:

• Multishot MRI offers faster, high-resolution imaging but is susceptible to motion artifacts.
• Existing motion correction methods are limited in effectiveness and computationally intensive.

Purpose of the Study:

• To develop a novel generative adversarial network (GAN)-based framework for effective motion correction in multishot MRI.
• To improve the robustness and reduce computational time of motion artifact correction.

Main Methods:

• A GAN-based conjugate gradient SENSE (CG-SENSE) reconstruction framework was proposed.
• CG-SENSE reconstructed images from motion-corrupted k-space data.
• The GAN framework was applied to correct motion artifacts.

Main Results:

• The proposed method demonstrated robustness across varying motion degrees, shot numbers, and encoding trajectories.
• Quantitative and qualitative analyses confirmed significant reduction in motion artifacts.
• Computational time was reduced several-fold compared to state-of-the-art techniques.

Conclusions:

• The proposed GAN-based CG-SENSE framework effectively corrects motion artifacts in multishot MRI.
• This approach offers a more robust and computationally efficient solution for motion correction.
• The findings have implications for improving diagnostic accuracy in MRI.