Motion Artifact Correction in MRI using GAN-Based Channel Attention Transformer
Journal
BioCAS 2023 - 2023 IEEE Biomedical Circuits and Systems Conference, Conference Proceedings
ISBN
9798350300260
Date Issued
2023-01-01
Author(s)
Abstract
Magnetic resonance imaging (MRI) is a widely used medical imaging technique that produces precise and detailed anatomical as well as functional data of the human body. However, motion artifacts caused by patient movement during the scan can degrade image quality, making diagnosis more difficult. In this paper, we propose a new approach to motion artifact correction in MRI based on Transformer architecture, a deep learning model that has achieved state-of-the-art results in natural language processing and computer vision tasks. Our model processes motion-corrupted MRI images hierarchically, utilizing channel self-attention mechanism to capture motion-blurred patterns. It is challenging to obtain the paired motion-corrupted and clear images simultaneously in real-world situations. To assess the effectiveness of our approach, we trained and tested the artifact correction model using both real and synthetic motion artifact MRI datasets. Overall, our proposed approach provides a promising direction for motion artifact correction in MRI and has the potential to improve the accuracy and reliability of MRI-based diagnoses.
Subjects
Generative Adversarial Network | Motion Artifact | Self-Attention | Transformer
Type
conference paper