Multi-channel GAN-based calibration-free diffusion-weighted liver imaging with simultaneous coil sensitivity...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Magnetic Resonance Imaging

Background:

• Parallel imaging in Diffusion-Weighted Imaging (DWI) can be affected by motion artifacts, particularly in abdominal scans, due to mismatches in coil calibration.
• Existing reconstruction methods may struggle with motion-induced artifacts, impacting image quality and diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate an iterative multichannel generative adversarial network (iMCGAN) for simultaneous sensitivity map estimation and calibration-free image reconstruction in DWI.
• To improve the robustness of DWI reconstruction against motion artifacts in abdominal imaging.

Main Methods:

• An iterative multichannel generative adversarial network (iMCGAN) framework was constructed for simultaneous sensitivity map estimation and calibration-free image reconstruction.
• The iMCGAN framework was applied to data from 106 healthy volunteers and 10 patients with tumors.
• Performance was evaluated by comparing iMCGAN with SAKE, ALOHA-net, and DeepcomplexMRI using metrics like PSNR, SSIM, RMSE, and ADC map histograms.

Main Results:

• The iMCGAN model significantly outperformed SAKE, ALOHA-net, and DeepcomplexMRI in terms of Peak Signal-to-Noise Ratio (PSNR) for b=800 DWI with a 4x acceleration factor.
• iMCGAN achieved a PSNR of 41.82 ± 2.14, compared to 17.38 ± 1.78 for SAKE, 20.43 ± 2.11 for ALOHA-net, and 39.78 ± 2.78 for DeepcomplexMRI.
• The proposed iMCGAN model effectively avoided ghosting artifacts common in SENSE reconstructions caused by sensitivity map mismatches due to motion.

Conclusions:

• The iMCGAN framework iteratively refines sensitivity maps and reconstructed images without requiring additional acquisitions.
• This approach significantly improves reconstructed image quality and alleviates aliasing artifacts in the presence of patient motion during DWI acquisition.
• iMCGAN offers a robust solution for motion-robust DWI reconstruction, particularly beneficial for abdominal imaging.