一种融合参考通道和深度学习的人体全身可移动磁共振成像设备主动EMI抑制方法  

Anactive EMI suppression method for human whole body mobile MRI device by integrating reference channels and deep learning

作　　者：张冠华 陆荣生[1] 吴正秀 胡晓凯 倪中华[1] ZHANG Guanhua;LU Rongsheng;WU Zhengxiu;HU Xiaokai;NI Zhonghua(School of Mechanictd Engineering,Southeast University,Nanjing 210096,China)

机构地区：[1]东南大学机械工程学院,南京210096

出　　处：《中国体视学与图像分析》2023年第4期427-436,共10页Chinese Journal of Stereology and Image Analysis

基　　金：国家自然科学基金面上项目(52075098)。

摘　　要：目的传统的磁共振成像设备通常采用被动屏蔽方式,如使用屏蔽房来抑制成像环境中的电磁干扰(electromagnetic interference,EMI),但这种方法难以用于有可移动式需求的场景下。本文提出了一种基于深度学习和参考通道的主动EMI抑制方法,以实现超低场磁共振设备的可移动和无屏蔽化。方法本文提出了一种基于卷积神经网络(convolutional neural network,CNN)和参考通道的EMI抑制方法,并提出了一种参考通道的评价方法,基于这种评价方法对参考通道进行了优化。结果本文的主动EMI抑制方法在水模图像中实现了71%的EMI抑制率,实现了图像信噪比2.3倍的提升。结论本文提出了一种基于深度学习的主动EMI抑制方法,该方法可以抑制磁共振图像中的EMI信号部分,提高图像的质量。Objective Traditional MRI devices typically use passive shielding methods such as shielded rooms for electromagnetic interference(EMI)suppression,but these methods are difficult to apply in scenarios with the need of mobility.This paper proposes an active EMI suppression method based on deep learning and reference channels to realize the mobility and remove shielding for ultra-low field MRI devices.Methods This paper proposes an active EMI suppression method based on a convolutional neural network(CNN)and reference channels together with an evaluation method for reference channels.The reference channels are optimized based on this evaluation method.Results The proposed active EMI suppression method achieved an EMI suppression rate of 71%in water model images,and boosted the image signal-lo-noise ratio(SNR)2.3 limes.Conclusions This article proposes an aclive EMI suppression method based on deep learning.It can suppress EMI in magnetic resonance images and EMI suppression;deep learning improve image quality.

关 键 词：超低场磁共振 EMI抑制 深度学习 

分 类 号：O482.532[理学—固体物理] TP391.41[理学—物理]