Motion-Informed Deep Learning for Brain MR Image Reconstruction Framework

Zhifeng Chen,Kamlesh Pawar,Kh Tohidul Islam,Himashi Peiris,Gary Egan,Zhaolin Chen

2024-05-28

Abstract:Motion artifacts in Magnetic Resonance Imaging (MRI) are one of the frequently occurring artifacts due to patient movements during scanning. Motion is estimated to be present in approximately 30% of clinical MRI scans; however, motion has not been explicitly modeled within deep learning image reconstruction models. Deep learning (DL) algorithms have been demonstrated to be effective for both the image reconstruction task and the motion correction task, but the two tasks are considered separately. The image reconstruction task involves removing undersampling artifacts such as noise and aliasing artifacts, whereas motion correction involves removing artifacts including blurring, ghosting, and ringing. In this work, we propose a novel method to simultaneously accelerate imaging and correct motion. This is achieved by integrating a motion module into the deep learning-based MRI reconstruction process, enabling real-time detection and correction of motion. We model motion as a tightly integrated auxiliary layer in the deep learning model during training, making the deep learning model 'motion-informed'. During inference, image reconstruction is performed from undersampled raw k-space data using a trained motion-informed DL model. Experimental results demonstrate that the proposed motion-informed deep learning image reconstruction network outperformed the conventional image reconstruction network for motion-degraded MRI datasets.

Image and Video Processing,Computer Vision and Pattern Recognition,Machine Learning,Medical Physics

What problem does this paper attempt to address?

This paper proposes a novel deep learning approach to address motion artifacts in magnetic resonance imaging (MRI). During the MRI scanning process, approximately 30% of clinical scans are affected by motion artifacts due to patient movement, which degrade the image quality. Although various motion correction methods exist, the application of deep learning for MRI reconstruction and motion correction is usually handled separately.

In this paper, the researchers propose an image reconstruction framework that combines motion detection and correction, integrating the motion module into the deep learning-based MRI reconstruction process. During the training phase, they simulate motion as an auxiliary layer of the deep learning model, enabling the model to detect and correct motion in real-time. During the inference phase, the trained motion-aware deep learning model is used to reconstruct images from undersampled raw k-space data.

Experimental results demonstrate that the proposed motion-aware deep learning image reconstruction network outperforms traditional image reconstruction networks for motion-degraded MRI datasets. The paper also conducts empirical studies, including fast MRI data and prospectively acquired motion data, to validate the effectiveness of the proposed method. Furthermore, ablation studies are performed to compare the performance of the model on datasets without motion data.

Through this approach, the researchers successfully enhance the detection and correction capability of motion artifacts, improving the quality and efficiency of MRI images, thereby contributing to the improvement of healthcare services. However, this method is currently limited to handling rigid motions, and further improvements are needed for non-rigid motion handling. Additionally, comprehensive validation on patient data has not yet been conducted.