Abstract #4439

Retrospective Motion Artifact Correction Using Refinement U-Nets with Wavelet Affine Transformations and Adaptive Multi-Loss Normalization

Ahmed Hassan¹, Mahmoud Yaser¹, Ibrahim Mohamed¹, Maha Medhat¹, Mohamed Ismail¹, Meena M. Makary^1,2, and Mohammed A. Al-masni³

¹Systems and Biomedical Engineering Department, Cairo University, Cairo, Egypt, ²Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Charlestown, MA, United States, ³Department of Artificial Intelligence and Data Science, Sejong University, Seoul, Korea, Republic of

Synopsis

Keywords: Artifacts, Artifacts, Motion Artifacts

Motivation: MRI scans have inherently lengthy acquisition times, making them susceptible to motion artifacts that can degrade AI performance and compromise clinical diagnoses accuracy.

Goal(s): Our goal was to incorporate wavelet transformations and adaptive multi-loss functions to optimize artifact correction and achieve more accurate, high-quality MRI images

Approach: Our method integrates wavelet transformations within a refinement U-Net architecture, combined with adaptive multi-loss normalization, to accurately address artifact correction using real motion patterns from motion-free MRI scans.

Results: Our method achieved significant improvements, raising SSIM from 76.85% to 92.92% and PSNR from 24.96 to 32.78, demonstrating effective artifact correction and surpassing other retrospective methods.

Impact: Correcting motion artifacts in MRI scans enhances image quality, making them more reliable for clinical diagnosis. Additionally, using this approach as a preprocessing step for tasks like registration and segmentation boosts model accuracy and supports improved diagnostic outcomes.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.