北京生物医学工程

基于描峰聚类的动态脑功能网络状态划分

Dynamic functional brain network divisionbased on density peak clustering

作者：崔园

单位：北京交通大学计算机与信息技术学院(北京 100044)

关键词：动态功能连接；静息态功能磁共振成像；描峰聚类；脑功能网络；降维

分类号：R318.04

出版年·卷·期（页码）：2019·38·6（575-582）

摘要：

目的近年来脑功能网络的动态属性分析已经成为脑功能研究的热点，脑功能网络状态划分则是脑功能网络动态属性分析的重要方面，目前国际上广泛采用的脑功能网络状态划分策略是k-means聚类算法，而k-means聚类算法存在两个缺陷。而描峰聚类（density peak clustering）算法能直观展现合理的类别数，从而有效解决k-means聚类中k值难以确定的问题。本文拟基于动态功能连接（dynamic functional connectivity, DFC）的脑功能网络状态划分，为脑功能网络划分探索新的模型。方法基于61位成年人静息态功能磁共振成像(resting state functional magnetic resonance image, rs-fMRI)数据，采用滑动窗口计算方法构建功能连接矩阵。基于多种距离度量使用多维尺度分析算法对其进行有效降维，通过描峰聚类算法进行脑功能网络状态划分，使用脑功能网络划分常用的状态模式图和聚类决策图进行结果的校验。结果基于余弦距离、相关系数以及Spearman等描述相似性的距离度量进行降维，得到的结果生理意义较为明确，且有效功能网络状态数为3~5。另外，脑区之间松散联系的网络状态比其他网络状态更频繁地发生。结论描峰聚类算法足以对个体脑功能连接随时间的动态波动进行状态划分，这可为脑功能网络划分研究提供新的思路。

Objective In recent years, the dynamic attribute analysis of functional brain network has become a hot topic in functional brain research. The network state division is an important aspect of network dynamic attribute analysis. At present, the internationally widely used network state division strategy is k-means clustering algorithm which has two defects. Density peak clustering can visually display the reasonable number of categories, which can effectively solve the problem that k-values in k-means clustering are difficult to determine. We intend to explore a new model for network state partitioning based on dynamic functional connectivity (DFC). Methods Based on resting state functional magnetic resonance image (rs-fMRI) data of 61 adults, a sliding window calculation method is used to construct a functional connectivity matrix. Multidimensional scale analysis algorithm is used to effectively reduce dimension based on multiple distance metrics, and network state division is performed by density peak clustering. State pattern graph and cluster decision graph are used to verify the results. Results Dimension reduction based on cosine distance, correlation coefficient and distance metrics describing similarity such as Spearman, the physiological significance of the results are clear, and the number of effective functional network states is 3 to 5. In addition, network states that are loosely connected between brain regions occur more frequently than other network states. Conclusions Brain network state can be divided by DFC characteristics, which can provide new ideas for functional brain network partition.

参考文献：

[1] Espinoza FA, Liu J, Ciarochi J, et al. Dynamic functional network connectivity in Huntington's disease and its associations with motor and cognitive measures[J]. Human Brain Mapping, 2019, 40(6): 1955-1968.

[2] Sanfratello L, Houck J, Calhoun VD. Dynamic functional network connectivity in schizophrenia with megnetoencephalography and functional magnetic resonance imaging: Do different time scales tell a different story?[J]. Brain Connectivity, 2019, 9(3): 251-262.

[3] Zhang Y, Wang C, Wang Y, et al. Investigating dynamic functional network patterns after propofol-induced loss of consciousness[J]. Clinical Neurophysiology, 2019, 130(3): 331-340.

[4] Allen EA, Damaraju E, Plis SM, et al. Tracking whole-brain connectivity dynamics in the resting state[J]. Cerebral Cortex, 2014, 24(3): 663-676.

[5] Tian L, Li Q, Wang C, et al. Changes in dynamic functional connections with aging[J]. NeuroImage, 2018, 172: 31-39.

[6] Xu H, Shen H, Wang L, et al. Impact of 36 h of total sleep deprivation on resting-state dynamic functional connectivity[J]. Brain Research, 2018, 1688: 22-32.

[7] Sala-Llonch R, Bartrés-Faz D, Junqué C. Reorganization of brain networks in aging: a review of functional connectivity studies[J]. Frontiers in Psychology, 2015, 6: 663-674.

[8] Hutchison RM, Womelsdorf T, Gati JS, et al. Resting-state networks show dynamic functional connectivity in awake humans and anesthetized macaques[J]. Human Brain Mapping, 2013, 34(9): 2154-2177.

[9] Ma S, Calhoun VD, Phlypo R, et al. Dynamic changes of spatial functional network connectivity in healthy individuals and schizophrenia patients using independent vector analysis[J]. NeuroImage, 2014, 90: 196-206.

[10] Naik S, Surampudi BR, Banerjee A, et al. Metastability in Senescence[J]. Trends in Cognitive Science, 2017, 21(7): 509-521.

[11] Nooner KB, Colcombe SJ, Tobe RH, et al. The NKI-rockland sample: a model for accelerating the pace of discovery science in psychiatry[J]. Frontiers in Neuroscience, 2012, 6: 152-163.

[12] Meier TB, Desphande AS, Vergun S, et al. Support vector machine classification and characterization of age-related reorganization of functional brain networks[J]. NeuroImage, 2012, 60(1): 601-613.

[13] Dosenbach NUF, Nardos B, Cohen AL, et al. Prediction of individual brain maturity using fMRI[J]. Science, 2010, 329(5997): 1358-1361.

[14] Challis E, Hurley P, Serra L, et al. Gaussian process classification of Alzheimer's disease and mild cognitive impairment from resting-state fMRI[J]. Neuroimage, 2015, 112: 232-243.

[15] Craddock RC, James GA, Holtzheimer PE 3rd, et al. A whole brain fMRI atlas generated via spatially constrained spectral clustering[J]. Human Brain Mapping, 2012, 33(8): 1914-1928.

[16] Kim J, Calhoun VD, Shim E, et al. Deep neural network with weight sparsity control and pre-training extracts hierarchical features and enhances classification performance: evidence from whole-brain resting-state functional connectivity patterns of schizophrenia[J]. NeuroImage, 2016, 124 (Pt A): 127-146.

[17] Rashid B, Arbabshirani MR, Damaraju E, et al. Classification of schizophrenia and bipolar patients using static and dynamic resting-state fMRI brain connectivity[J]. NeuroImage, 2016, 134: 645-657.

[18] Zalesky A, Breakspear M. Towards a statistical test for functional connectivity dynamics[J]. NeuroImage, 2015, 114: 466-470.

[19] Rodriguez A, Laio A. Machine learning. Clustering by fast search and find of density peaks[J]. Science, 2014, 344(6191): 1492-1496.

[20] 于剑. 机器学习：从公理到算法[M]. 北京：清华大学出版社, 2017.

[21] 付先凯, 蒋鑫龙, 刘军发, 等. 基于多维尺度分析的自适应室内群终端定位方法[J]. 计算机科学, 2018, 45(10): 111-117.

Fu XK, Jiang XL, Liu JF, et al. adaptive indoor location method for multiple terminals based on multidimentional scaling[J]. Computer Science, 2018, 45(10): 111-117.

[22] 周志华. 机器学习[M]. 北京：清华大学出版社, 2016.

Zhou ZH. Machine learning[M]. Beijing：Tsinghua University Press, 2016.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录