-
健康危险行为是对人们的身体健康和精神状态产生危害或潜在危害的行为[1],与情感和认知有密切关系。尤其是青春期的个体,由于身心发展不平衡,更容易卷入危害健康的行为之中,但目前对青少年健康危险行为的研究主要是行为学的研究。文献[2]对6 633名青少年的研究中发现,在所有的危险行为上有较高概率的高危险组占全体人员的13.6%,这对青少年的身心健康乃至成年后的生活造成了严重的伤害。实证研究发现,不同形式的健康危险行为(比如暴力攻击、吸毒、违反纪律等)常常伴随发生[3],为了解释这种现象,文献[4]提出了危险行为理论,认为各种不同形式的危险行为是由潜在的共同问题行为因子决定的。这种潜在的问题行为因子是否跟大脑的网络组织有关呢?可能跟哪些脑区有关,脑区间有怎样的相互作用,以及又有怎样的脑网络特征?本文拟通过利用动态功能连接对健康危险性行为特征的预测来研究与青少年健康危险行为相关的脑网络模式。
在目前有关功能连接的研究中,功能连接主要是通过计算在整个静息态fMRI扫描时程中各脑区的时间序列相关性得到,称之为静态功能连接[5]。然而最近研究表明在静息态下的功能连接也是动态变化的,显示出显著的振幅波动性[6],而且在静息态下观测到的低频振荡呈现出复杂的空时结构,包含多个离散的稳定模式[7]。因此,研究认为静息态下功能连接的动态性反应了与认知和行为能力相关的脑区间动态的相互作用[8]。猜想这种动态的相互作用可能与青少年的健康危险行为相关。因此,本文的研究通过使用基于支持向量机(SVM)的回归分析(SVR)利用动态功能连接对个体的健康危险行为评分进行预测,以实现对青少年健康危险行为的脑网络研究。
-
作为第一步,本文评估了全脑连接是否能预测晚期青少年的健康危险性行为。基于特征选择阈值P < 0.01,预测和观察值之间的相关系数r=0.388 2 (P = 0.005 8),产生了显著的预测,其他阈值P < 0.008, 0.015下也有显著性,本文选取P < 0.01进行分析[14]。
对正的特征模型,基于特征选择阈值P < 0.01,预测和观察值之间的相关系数r=0.426 2 (P = 0.002 3),对负的特征模型在多项式核函数下没有产生显著的预测,但是负的特征模型在线性核函数下显示出了显著性r=0.446 2 (P =0.001 3)。
-
由于留一法交叉验证中,每一次迭代中所选择的训练样本集略有差异,当P < 0.01时,出现在每一轮交叉验证里的一致性功能连接一共有31条(如表 1、图 1所示),可见这31条动态功能连接具有较强的预测能力,其中14条功能连接与健康危险性行为呈正相关,17条功能连接与健康危险性行为呈负相关。
表 1 与HBICA相关的连接
功能连接 MNI坐标 r p 与HBICA正相关的连接 'dFC/med_cerebellum' [-42,7,36]/[-16,-64,-21] 0.578 1 0.000 0 'post_parietal/post_occipital' [-41,-31,48]/[33,-81,-2] 0.530 4 0.000 1 post_occipital/vmPFC' [-29,-88,8]/[6,64,3] 0.529 5 0.000 1 'post_parietal/post_occipital' [-41,-31,48]/[-29,-88,8] 0.515 4 0.000 2 'mid_insula/ACC' [-42,-3,11]/[9,39,20] 0.485 5 0.000 5 'mid_insula/dlPFC' [-42,-3,11]/[46,28,31] 0.480 5 0.000 6 'vFC/vmPFC' [43,1,12]/[-11,45,17] 0.470 0 0.000 9 'parietal/IPL' [-24,-30,64]/[-48,-47,49] 0.453 6 0.001 4 'post_insula/vmPFC' [42,-24,17]/[-11,45,17] 0.452 8 0.001 4 'precentral_gyrus/aPFC' [-54,-9,23]/[-29,57,10] 0.449 6 0.001 6 'parietal/aPFC' [-24,-30,64]/[-29,57,10] 0.448 8 0.001 5 'parietal/IPL' [-24,-30,64]/[-41,-40,42] 0.445 3 0.001 7 'angular_gyrus/basal_ganglia' [-41,-47,29]/[-6,17,34] 0.429 3 0.002 5 'mid_insula/ACC' [33,-12,16]/[-2,30,27] 0.420 7 0.003 1 与HBICA负相关的连接 'post_cingulate/vFC' [-4,-31,-4]/[-48,6,1] -0.515 1 0.000 2 'precentral_gyrus/post_cingulate' [46,-8,24]/[-4,-31,-4] -0.505 2 0.000 3 'IPL/post_cingulate' [-53,-50,39]/[-4,-31,-4] -0.502 0 0.000 3 'precentral_gyrus/thalamus' [-44,-6,49]/[11,-12,6] -0.499 8 0.000 3 'precentral_gyrus/med_cerebellum' [-44,-6,49]/[-11,-72,-14] -0.471 6 0.000 8 'occipital/temporal' [-9,-72,41]/[51,-30,5] -0.469 5 0.000 8 'temporal/occipital' [59,-13,8]/[-9,-72,41] -0.458 4 0.001 2 'post_cingulate/ACC' [10,-55,17]/[9,39,20] -0.439 4 0.002 0 'IPL/aPFC' [54,-44,43]/[-29,57,10] -0.439 3 0.002 0 'precentral_gyrus/thalamus' [-44,-6,49]/[-12,-12,6] -0.434 7 0.002 2 'dFC/TPJ' [-42,7,36]/[-52,-63,15] -0.429 8 0.002 5 'precuneus/occipital' [5,-50,33]/[-28,-42,-11] -0.428 0 0.002 5 'post_insula/thalamus' [42,-24,17]/[-12,-12,6] -0.426 2 0.002 7 'IPL/thalamus' [-53,-50,39]/[-12,-12,6] -0.423 6 0.002 9 'aPFC/med_cerebellum' [27,49,26]/[14,-75,-21] -0.423 5 0.002 9 'precuneus/vlPFC' [9,-43,25]/[46,39,-15] -0.422 6 0.002 9 'precentral_gyrus/occipital' [44,-11,38]/[-28,-42,-11] -0.414 7 0.003 6 为了考察各个网络对健康危险性行为的预测能力情况,本文计算了这些网络内边和网络间边的权重。对每个网络的特征权重计算结果分析发现感觉运动网络,带状盖网络,默认网络和额顶网络相对有较大的总权重,意味着这4个网络对健康危险性行为有较大的影响,如图 2所示。按照正负相关的进一步分析发现在正相关中感觉运动网络相关性最强,而负相关中带状盖网络,默认网络和额顶网络有较强的预测能力。
基于功能连接是属于网络内部的连接还是网络之间的连接进行分类,发现在默认网络内部和带状盖网络内部的连接相对较丰富,且与健康危险性行为呈负相关。其他网络内部连接都比较少,绝大部分连接位于网络之间,且主要呈现为带状盖网络和额顶网络之间的连接,以及感觉运动网络与带状盖网络、默认网络和额顶网络之间的连接呈现出了异常(见图 2和表 1)。
-
不同脑区对预测的贡献大小不同,有些脑区发挥的作用大,有些脑区发挥的作用小。本文利用脑区加权来标示脑区对预测所起的作用,即对所有跟该脑区相连的一致性功能连接进行权重求和并除以2。与一致性功能连接相关的脑区如图 1所示,球的直径大小表示了脑区权重,直径越大,预测能力越强。由图可知几个脑区相比其他脑区直径更大,表明它们具有较强的预测能力,主要是位于带状盖网络的扣带后回(post_cingulate)、丘脑(thalamus)、前额叶腹部(vFC),位于感觉运动网络的中央前回(precentral_gyrus)、顶叶中部及后部(parietal,post_parietal)、脑岛(insula),位于额顶控制网络的前额叶前部(aPFC),前额叶背侧(dFC),顶下小叶(IPL),位于默认网络的前扣带回(ACC),腹内侧前额叶皮层(vmPFC)。
Prediction of the Health-Risk Behavior by Using Dynamic Functional Connectivity
-
摘要: 为了研究健康危险性行为的脑网络特征,该文采集了49个被试的静息态功能磁共振数据。使用每一个对象动态功能连接网络的低频振荡振幅作为特征,利用支持向量回归对个体的健康危险行为进行预测。结果表明动态功能连接能较好地预测健康危险性行为特征,并提取了与之相关的功能连接模式,对预测有重要作用的连接绝大部分位于网络之间,且主要呈现为带状盖网络和额顶网络之间的连接,以及感觉运动网络与它们之间的连接相关。Abstract: In order to investigate the brain network characteristics of the health-risk behavior, we collected fMRI data of 49 subjects under rest state. The fluctuation amplitude of dynamic functional connectivity is used as the features of support vector regression (SVR) to predict the health-risk behavior. The results show a good correlation between spontaneous fluctuation of rest state and the health-risk behavior. Some informational functional connectivities could be used to predict the health-risk behavior and they mainly locate among the connections of networks:mainly cingulo-opercular network, frontoparietal network, sensorimotor network, etc..
-
表 1 与HBICA相关的连接
功能连接 MNI坐标 r p 与HBICA正相关的连接 'dFC/med_cerebellum' [-42,7,36]/[-16,-64,-21] 0.578 1 0.000 0 'post_parietal/post_occipital' [-41,-31,48]/[33,-81,-2] 0.530 4 0.000 1 post_occipital/vmPFC' [-29,-88,8]/[6,64,3] 0.529 5 0.000 1 'post_parietal/post_occipital' [-41,-31,48]/[-29,-88,8] 0.515 4 0.000 2 'mid_insula/ACC' [-42,-3,11]/[9,39,20] 0.485 5 0.000 5 'mid_insula/dlPFC' [-42,-3,11]/[46,28,31] 0.480 5 0.000 6 'vFC/vmPFC' [43,1,12]/[-11,45,17] 0.470 0 0.000 9 'parietal/IPL' [-24,-30,64]/[-48,-47,49] 0.453 6 0.001 4 'post_insula/vmPFC' [42,-24,17]/[-11,45,17] 0.452 8 0.001 4 'precentral_gyrus/aPFC' [-54,-9,23]/[-29,57,10] 0.449 6 0.001 6 'parietal/aPFC' [-24,-30,64]/[-29,57,10] 0.448 8 0.001 5 'parietal/IPL' [-24,-30,64]/[-41,-40,42] 0.445 3 0.001 7 'angular_gyrus/basal_ganglia' [-41,-47,29]/[-6,17,34] 0.429 3 0.002 5 'mid_insula/ACC' [33,-12,16]/[-2,30,27] 0.420 7 0.003 1 与HBICA负相关的连接 'post_cingulate/vFC' [-4,-31,-4]/[-48,6,1] -0.515 1 0.000 2 'precentral_gyrus/post_cingulate' [46,-8,24]/[-4,-31,-4] -0.505 2 0.000 3 'IPL/post_cingulate' [-53,-50,39]/[-4,-31,-4] -0.502 0 0.000 3 'precentral_gyrus/thalamus' [-44,-6,49]/[11,-12,6] -0.499 8 0.000 3 'precentral_gyrus/med_cerebellum' [-44,-6,49]/[-11,-72,-14] -0.471 6 0.000 8 'occipital/temporal' [-9,-72,41]/[51,-30,5] -0.469 5 0.000 8 'temporal/occipital' [59,-13,8]/[-9,-72,41] -0.458 4 0.001 2 'post_cingulate/ACC' [10,-55,17]/[9,39,20] -0.439 4 0.002 0 'IPL/aPFC' [54,-44,43]/[-29,57,10] -0.439 3 0.002 0 'precentral_gyrus/thalamus' [-44,-6,49]/[-12,-12,6] -0.434 7 0.002 2 'dFC/TPJ' [-42,7,36]/[-52,-63,15] -0.429 8 0.002 5 'precuneus/occipital' [5,-50,33]/[-28,-42,-11] -0.428 0 0.002 5 'post_insula/thalamus' [42,-24,17]/[-12,-12,6] -0.426 2 0.002 7 'IPL/thalamus' [-53,-50,39]/[-12,-12,6] -0.423 6 0.002 9 'aPFC/med_cerebellum' [27,49,26]/[14,-75,-21] -0.423 5 0.002 9 'precuneus/vlPFC' [9,-43,25]/[46,39,-15] -0.422 6 0.002 9 'precentral_gyrus/occipital' [44,-11,38]/[-28,-42,-11] -0.414 7 0.003 6 -
[1] 季成叶.青少年健康危险行为[J].中国学校卫生, 2007, 28(4):289-291. doi: 10.3969/j.issn.1000-9817.2007.04.001 JI Chen-ye. The health-risk behavior of adolescents[J].Chinese Journal of School Health, 2007, 28(4):289-291. doi: 10.3969/j.issn.1000-9817.2007.04.001 [2] WANG M, YI J, CAI L, et al. Development and psychometric properties of the health-risk behavior inventory for Chinese adolescents[J]. BMC Med Res Methodol, 2012, 12:94. doi: 10.1186/1471-2288-12-94 [3] WELTE J W, WIECZOREK W F, BARNES G M, et al. The relationship of ecological and geographic factors to gambling behavior and pathology[J]. J Gambl Stud, 2004, 20(4):405-423. doi: 10.1007/s10899-004-4582-y [4] JESSOR R. Risk behavior in adolescence:a psychosocial framework for understanding and action[J]. Developmental Review, 1992, 12(4):374-390. doi: 10.1016/0273-2297(92)90014-S [5] JIANG W, SHI F, LIAO J, et al. Disrupted functional connectome in antisocial personality disorder[J]. Brain Imaging Behav, 2016, doi: 10.1007/s11682-016-9572-z. [6] HUTCHISON R M, WOMELSDORF T, GATI J S, et al. Resting-state networks show dynamic functional connectivity in awake humans and anesthetized macaques[J]. Hum Brain Mapp, 2013, 34(9):2154-2177. doi: 10.1002/hbm.v34.9 [7] ALLEN E A, DAMARAJU E, PLIS S M, et al. Tracking whole-brain connectivity dynamics in the resting state[J]. Cereb Cortex, 2014, 24(3):663-676. doi: 10.1093/cercor/bhs352 [8] HUTCHISON R M, MORTON J B. Tracking the brain's functional coupling dynamics over development[J]. J Neurosci, 2015, 35(17):6849-6859. doi: 10.1523/JNEUROSCI.4638-14.2015 [9] SHEN H, WANG L, LIU Y, et al. Discriminative analysis of resting-state functional connectivity patterns of schizophrenia using low dimensional embedding of fMRI[J]. NeuroImage, 2010, 49:3110-3121. doi: 10.1016/j.neuroimage.2009.11.011 [10] DOSENBACH N U, NARDOS B, COHEN A L, et al. Prediction of individual brain maturity using fMRI[J]. Science, 2010, 329(5997):1358-1361. doi: 10.1126/science.1194144 [11] YU R, CHIEN Y L, WANG H L, et al. Frequency-specific alternations in the amplitude of low-frequency fluctuations in schizophrenia[J]. Hum Brain Mapp, 2014, 35(2):627-637. doi: 10.1002/hbm.v35.2 [12] LEONARDI N, VAN DE VILLE D. On spurious and real fluctuations of dynamic functional connectivity during rest[J]. Neuroimage, 2015, 104:430-436. doi: 10.1016/j.neuroimage.2014.09.007 [13] SHEN H, LI Z, QIN J, et al. Changes in functional connectivity dynamics associated with vigilance network in taxi drivers[J]. Neuroimage, 2016, 124:367-378. doi: 10.1016/j.neuroimage.2015.09.010 [14] FINN E S, SHEN X, SCHEINOST D, et al. Functional connectome fingerprinting:identifying individuals using patterns of brain connectivity[J]. Nat Neurosci, 2015, 18(11):1664-1671. doi: 10.1038/nn.4135 [15] QIN J, CHEN S G, HU D, et al. Predicting individual brain maturity using dynamic functional connectivity[J]. Front Hum Neurosci, 2015, 9:418. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=Doaj000004205449 [16] FAIR D A, DOSENBACH N U, CHURCH J A, et al. Development of distinct control networks through segregation and integration[J]. Proc Natl Acad Sci USA, 2007, 104(33):13507-13512. doi: 10.1073/pnas.0705843104 [17] SYLVESTER C M, CORBETTA M, RAICHLE M E, et al. Functional network dysfunction in anxiety and anxiety disorders[J]. Trends Neurosci, 2012, 35(9):527-535. doi: 10.1016/j.tins.2012.04.012 [18] BUCKNER R L, ANDREWS-HANNA J R, SCHACTER D L. The brain's default network:anatomy, function, and relevance to disease[J]. Ann N Y Acad Sci, 2008, 1124:1-38. doi: 10.1196/annals.1440.011