Fetal MRI Artifacts: Semi-Supervised Generative Adversarial Neural Network for Motion Artifacts Reducing in Fetal Magnetic Resonance Images  

作　　者：Ítalo Messias Félix Santos Gilson Antonio Giraldi Heron Werner Junior Bruno Richard Schulze Ítalo Messias Félix Santos;Gilson Antonio Giraldi;Heron Werner Junior;Bruno Richard Schulze(National Laboratory for Scientific Computing, Rio de Janeiro, Brazil;Biodesign Laboratory Dasa/PUC-Rio, Rio de Janeiro, Brazil)

机构地区：[1]National Laboratory for Scientific Computing, Rio de Janeiro, Brazil [2]Biodesign Laboratory Dasa/PUC-Rio, Rio de Janeiro, Brazil

出　　处：《Journal of Computer and Communications》2024年第6期210-225,共16页电脑和通信（英文）

摘　　要：This study addresses challenges in fetal magnetic resonance imaging (MRI) related to motion artifacts, maternal respiration, and hardware limitations. To enhance MRI quality, we employ deep learning techniques, specifically utilizing Cycle GAN. Synthetic pairs of images, simulating artifacts in fetal MRI, are generated to train the model. Our primary contribution is the use of Cycle GAN for fetal MRI restoration, augmented by artificially corrupted data. We compare three approaches (supervised Cycle GAN, Pix2Pix, and Mobile Unet) for artifact removal. Experimental results demonstrate that the proposed supervised Cycle GAN effectively removes artifacts while preserving image details, as validated through Structural Similarity Index Measure (SSIM) and normalized Mean Absolute Error (MAE). The method proves comparable to alternatives but avoids the generation of spurious regions, which is crucial for medical accuracy.This study addresses challenges in fetal magnetic resonance imaging (MRI) related to motion artifacts, maternal respiration, and hardware limitations. To enhance MRI quality, we employ deep learning techniques, specifically utilizing Cycle GAN. Synthetic pairs of images, simulating artifacts in fetal MRI, are generated to train the model. Our primary contribution is the use of Cycle GAN for fetal MRI restoration, augmented by artificially corrupted data. We compare three approaches (supervised Cycle GAN, Pix2Pix, and Mobile Unet) for artifact removal. Experimental results demonstrate that the proposed supervised Cycle GAN effectively removes artifacts while preserving image details, as validated through Structural Similarity Index Measure (SSIM) and normalized Mean Absolute Error (MAE). The method proves comparable to alternatives but avoids the generation of spurious regions, which is crucial for medical accuracy.

关 键 词：Fetal MRI Artifacts Removal Deep Learning Image Processing Generative Adversarial Networks 

分 类 号：TN3[电子电信—物理电子学]