Deep Learning With U-Net for Motion Artifact Reduction in Brain MRI

Abstract

Magnetic Resonance Imaging (MRI) provides detailed three-dimensional visualization of internal structures and plays a crucial role in diagnosing neurological disorders. However, motion artifacts caused by patient movement often degrade image quality and hinder accurate diagnosis. In this study, we propose a U-Net–based deep learning approach to reduce motion artifacts in brain MRI. Standard images without artifacts and motion-corrupted images were obtained from the OpenNeuro database and used to train and evaluate the model. The network was optimized with a fixed learning rate of 0.0001, and performance was assessed using peak signal-to-noise ratio (PSNR) and loss metrics. Experimental results demonstrated a reduction in the loss value from 1.0615 to 0.4424 and an improvement in PSNR from 5.76 to 9.56 after 100 epochs. These findings indicate that the proposed method effectively suppresses motion artifacts and reconstructs brain MRI images with improved diagnostic quality.