Estimation of independent component numbers based on improved effective detection criteria for resting state functional magnetic resonance imaging(rfMRI)of Alzheimers disease(AD)

doi:10.3969/j.issn.1000-1565.2020.03.013

Abstract

Abstract: The prevalence of Alzheimers disease(AD)has been increasing year after year in the elderly. Functional connection analysis based on resting state functional magnetic resonance imaging(rfMRI)is an important method to study the pathogenesis of AD. In order to extract the region of interest of rfMRI image and analyze the change of functional connection effectively, it is necessary to separate independent components.- DOI:10.3969/j.issn.1000-1565.2020.03.013AD-rfMRI图像的基于改进EDC的独立分量数目估计郑伟^1,2,3,王轩^1,2,3,姚纪智^1,2,3,刘帅奇^1,2,3,张晓丹⁴,马泽鹏⁴(1.河北大学电子信息工程学院,河北保定 071002;2.河北省数字医疗工程重点实验室, 河北保定 071002;3.河北省机器视觉工程技术研究中心, 河北保定 071002;4.河北大学附属医院 CT-MRI诊断科, 河北保定 071000)摘要:阿尔茨海默病(Alzheimers disease,AD)在老年人中的患病人数逐年升高,而静息态功能磁共振成像(rfMRI)功能连接变化是研究AD发病机制的重要手段.为了有效地提取静息态功能磁共振成像图像感兴趣区域并进行功能连接变化分析,需要分离独立分量.而独立分量的分离实现,必须预先估计需要分离的独立分量的数目.在独立分量数目估计方法中,信息理论准则中的有效检测准则(effective detection criteria,EDC)具有灵活的惩罚函数,估计结果鲁棒性高,但存在过估计问题,影响独立分量分离结果的准确性.本文将对数函数引入惩罚函数的第2种表示(EDC₂)对其进行改进,并结合黄金分割法确定惩罚函数项的最优值,称为OIEDC₂(optimizing and improving effective detection criteria).实验结果表明,OIEDC₂与EDC₂相比提升了独立分量数目估计的合理性和准确性.关键词:阿尔茨海默病;静息态功能磁共振成像;独立分量分析;有效检测准则中图分类号:TN911 文献标志码:A 文章编号:1000-1565(2020)03-0315-07Estimation of independent component numbers based on improved effective detection criteria for resting state functional magnetic resonance imaging(rfMRI)of Alzheimers disease(AD)ZHENG Wei^1,2,3, WANG Xuan^1,2,3, YAO Jizhi^1,2,3, LIU Shuaiqi^1,2,3, ZHANG Xiaodan⁴, MA Zepeng⁴(1. College of Electronic and Information Engineering, Hebei University, Baoding 071002, China; 2. Machine Vision Engineering Research Center of Hebei Province, Baoding 071002, China; 3. Key Laboratory of Digital Medical Engineering of Hebei Province, Baoding 071002, China; 4. Department of CT-MRI, Affiliated Hospital of Hebei University,Baoding 071000, China)Abstract: The prevalence of Alzheimers disease(AD)has been increasing year after year in the elderly. Functional connection analysis based on resting state functional magnetic resonance imaging(rfMRI)is an important method to study the pathogenesis of AD. In order to extract the region of interest of rfMRI image and analyze the change of functional connection effectively, it is necessary to separate independent components.- 收稿日期:2019-05-01 基金项目:国家自然科学基金资助项目(61401308, 61572063);河北省自然科学基金资助项目( F2018210148, F2016201142, F2016201187);河北省教育厅项目(QN2016085); 河北省机器视觉工程技术研究中心开放课题(2018HBMV02);河北大学引进人才科研启动经费资助项目(2014-303) 第一作者:郑伟(1972—),女,黑龙江兰西人,河北大学教授,博士,主要从事图像处理与分析、图像安全通信、图像加密和隐藏研究. E-mail: 147685650@qq.com 通信作者:马泽鹏(1989—),男,河北保定人,河北大学附属医院CT-MRI诊断科医师,主要从事脑部和心脏疾病的CT、MR、fMRI的诊断及鉴别诊断. E-mail: mzpdan@163.com第3期郑伟等:AD-rfMRI图像的基于改进EDC的独立分量数目估计In order to separate independent components, the number of independent components to be separated must be estimated in advance. In the estimation of the number of independent components, the effective detection criteria(EDC)in the information theory criteria has a flexible penalty function and high robustness, but there are over estimation problems, which affect the accuracy of the independent component separation results. Here, Logarithm function is introduced into penalty function to improve EDC₂, and the optimal value of penalty function term is determined by golden section method, which is called OIEDC₂. Experimental results show that compared with EDC₂, OIEDC₂ improves the rationality and accuracy of independent component number estimation.

Key words: Alzheimers disease, independent component analysis, resting state functional magnetic resonance imaging, effective detection criteria

CLC Number:

TN911

ZHENG Wei, WANG Xuan, YAO Jizhi, LIU Shuaiqi, ZHANG Xiaodan, MA Zepeng. Estimation of independent component numbers based on improved effective detection criteria for resting state functional magnetic resonance imaging(rfMRI)of Alzheimers disease(AD)[J]. Journal of Hebei University (Natural Science Edition), 2020, 40(3): 315-321.

References

[1] ZHANG S, LI C R. Functional connectivity parcellation of the human thalamus by independent component analysis[J]. Brain Connectivity, 2017, 7(9): 602-616. DOI: 10.1089/brain.2017.0500.
[2] LIU S, ZHAO C, AN Y, et al. Diffusion tensor imaging denoising based on Riemannian geometric framework and sparse Bayesian learning[J]. Journal of Medical Imaging and Health Informatics, 2019, 9(9): 1993-2003. DOI: 10.1166/jmihi.2019.2832.
[3] 刘怡秋, 孙曼, 陈元园, 等. 静息态fMRI观察轻中度阿尔茨海默病患者脑区功能连接异常[J]. 中国医学影像技术, 2018, 34(09): 53-57. DOI: 10.13929/j.1003-3289.201803187.
[4] 郑伟, 王轩, 刘敬, 等. 女性阿尔茨海默病患者静息态功能磁共振图像FC分析[J]. 河北大学学报(自然科学版), 2018, 38(5): 533-542. DOI: 0.3969/j.issn.1000-1565.2018.05.014.
[5] 马文洁, 赵小虎, 王湘彬, 等. 阿尔茨海默病患者脑网络间交互作用异常的静息fMRI研究[J]. 同济大学学报(医学版), 2019, 40(1): 68-72. DOI: 10. 16118 / j. 1008-0392.2019.01.013.
[6] ANDĚL, J, PEREZ M G, NEGRAO A I. Estimating the dimension of a linear model[J]. Kybernetika Praha, 1981, 17(6): 514-525. DOI: 10.1016/0047-259x(81)90089-0.
[7] STORTI S F, FORMAGGIO E, ACLER M, et al. Selection of optimal hemodynamic response function for fMRI analysis on acute stroke patients[J]. Neuroimage, 2010, 25(4): 253-256. DOI: 10.1016/S1053-8119(09)70524-5.
[8] RISSANEN J. Modelling by the shortest data description[J]. Automatica, 1978, 14(5): 465-471. DOI: 10.1016/0005-1098(78)90005-5.
[9] CAVANAUGH J E. A large-sample model selection criterion based on Kullbacks symmetric divergence[J]. Statistics & Probability Letters, 1999, 42(4): 333-343. DOI: 10.1016/S0167-7152(98)00200-4.
[10] 陈军, 石玉虎, 王倪传, 等. 基于有效检测准则的fMRI源信号数目的估计方法[J]. 安徽大学学报(自然科学版), 2016, 40(1): 64-72. DOI: 10.3969/j.issn.1000-2162.2016.01.011.