基于非局部均值与线性最小均方误差估计的MRI去噪研究  

Study on MRI Denoising Based on NLM and LMMSE Estimation

作　　者：吴娟 荆斌 荆钧尧 吴斌 孙娜娜 WU Juan;JING Bin;JING Junyao;WU Bin;SUN Nana(Department of Information,Tangdu Hospital,Airforce Medical University,Xi’an Shaanxi 710032,China;Department of Medical Engineering,Medical Supplies Center,Chinese PLA General Hospital,Beijing 100853,China;Department of Technology Development,North Electronics Research Institute,Xi’an Shaanxi 710100,China)

机构地区：[1]空军军医大学(第四军医大学)唐都医院信息科,陕西西安710032 [2]解放军总医院医疗保障中心医学工程科,北京100853 [3]北方电子研究所科技发展部,陕西西安710100

出　　处：《中国医疗设备》2025年第2期35-39,66,共6页China Medical Devices

摘　　要：目的提出一种针对磁共振成像(Magnetic Resonance Imaging,MRI)图像的Rician噪声去除算法。方法首先利用局部方差统计估计MRI的噪声水平,接着采用线性最小均方误差估计及非局部均值滤波方法对图像进行复原,再根据估计的图像噪声水平决定是否进行迭代去噪。结果利用模拟的大脑MRI对提出的去噪方法进行定性与定量验证。结果显示,去噪算法在噪声方差为15时,不同切片的均方误差、峰值信噪比与信噪比平均值依次为70.07、29.78 dB、21.95 dB,非局部均值滤波的结果依次为82.17、29.11 dB、21.28 dB,而线性最小均方误差估计的结果依次为108.16、27.80dB、19.97dB,可以看出本文提出的算法优于其他算法。相比传统的非局部均值滤波,本文提出的算法在边缘等信息保护方面也有一定提高,同时提高了线性最小均方误差估计在高噪声水平时的去噪效果。结论本文提出的算法能够有效实现含噪MRI信号的复原,为后续图像处理及应用提供可靠保证。Objective To propose a Rician noise removal algorithm for magnetic resonance imaging(MRI).Methods Firstly,the noise level of MRI was estimated by local variance statistics,and then the image was restored by linear least mean square error estimation and non-local means filtering.Results The proposed denoising method was verified qualitatively and quantitatively by using simulated brain MRI.The results showed that when the noise variance of the denoising algorithm was 15,the mean square error,peak signal-to-noise ratio and mean signal-to-noise ratio of different slices were 70.07,29.78 dB and 21.95 dB successively,and the results of non-local means filtering were 82.17,29.11 dB and 21.28 dB successively.The results of linear least mean square error estimation were 108.16,27.80 dB and 19.97 dB successively.It could be seen that the proposed algorithm was superior to other algorithms.Compared with the traditional non-local means filtering,the proposed algorithm also had a certain improvement in edge protection and improved the denoising effect of linear least mean square error estimation at high noise levels.Conclusion The algorithm proposed in this paper can effectively realize the restoration of noisy MRI signal and provide a reliable guarantee for the subsequent image processing and application.

关 键 词：磁共振成像(MRI) 去噪 非局部均值 线性最小均方误差 Rician噪声 自适应 迭代 

分 类 号：R197.39[医药卫生—卫生事业管理] TP391[医药卫生—公共卫生与预防医学]