认知与发展心理学研究室知识库

生物多样性的表现之一是个体差异，对其神经机制的研究是心理学和神经科学的前沿方向，核心与关键是对个体差异的可信测量。类内相关系数(intra-class correlation, ICC)是一种常被应用于考察个体差异的可信度统计量。磁共振技术通过非侵入性的方式对大脑的结构和功能变化进行成像(magnetic resonance imaging, MRI)，己成为心理学及神经科学领域最为倚重的研究工具之一。己有多项研究系统地考察了这些人脑结构形态和功能MRI测量的重测信度，但鲜有研究考察个体间差异的测量是否受到成像仪器、扫描序列等的影响。本研究首先建立多站点重复测量大型数据集，以人脑大尺度功能网络为研究对象，通过对其结构形态和内在功能指标的多层混合线性建模，定量化研究个体间差异、个体内差异和影像站点差异以及它们对测量信度的影响。包括以下三个研究:

研究一:介绍R3 BRAIN数据集及其统计学模型，共51名被试在2个站点完成4次扫描，建立混合线性模型，考察影响其人脑形态学变化的各类因素。结果表明:脑网络体积的个体间差异受站点因素影响显著，主要表现在其个体皮层厚度差异受站点的影响较大，而其个体皮层表面积差异则受站点影响较小，脑网络形态测量站点内重测信度(ICC＞0.8)高于站点间重测信度(ICC＞0.6)。

研究二:基于研究一的线性混合模型，采取双回归算法，分别考察了大尺度脑网络的自发低频波动振幅，网络间功能连接和网络内功能连接的个体差异。结果表明:回归全脑平均信号在代表性时间序列的振幅及功能连接两指标上受站点差异影响较小，双回归脑网络不同方法没有表现出差异。总体上控制网络和默认网络等高级网络表现出了更明显的个体差异。各网络受站间差异的影响主要表现在高重测信度的区域减少和重测信度高于0.4的区域减少两方面。

研究三:通过个体重复测量的线性混合模型，考察人脑大尺度网络形态和功能的性别效应。结果表明:性别相关的效应集中表现在全局特性，而这类全局差异得到适当控制后，男女之间的脑网络差异不再显著;具体来讲，男女形态学差异主要表现为男性脑容量较大，并且主要受皮层表面积差异的影响，这一男女差异也表现在脑网络水平上;男性在控制网络、注意网络和体感运动网络间具有更强的功能连接，体感运动网络内在中央沟区域的功能连接也高于女性。上述男女差别在相应控制全局特性后不再显著，并且在不同站点具备可重复性。

总结起来，本研究为多站点研究提供了参考。发现了脑网络形态个体差异测量具有近乎完美的重测信度，而功能个体差异的测量表现出中等到可观的重测信度，成像测量的信噪比是影响重测信度的重要因素。脑网络形态的个体差异测量具有相当可观到近乎完美的站点间重测信度，功能个体差异测量表现出了网络依赖的站点信度，控制网络、背侧注意网络、默认网络等具有中等的站点信度，同时信噪比也是对站点信度造成影响的重要因素。顶叶记忆网络的皮层形态具备很高的个体差异，联合皮层网络比初级皮层网络在自发功能活动上具有更高的个体间的差异，而脑网络性别差异主要受结构和功能的全局指标驱动。

Individual differences is one of the most important features of biodiversity and the study of its neural mechanisms is the frontier of psychology and neuroscience. The key point of research individual differences is its stable measurement. Intra-class correlation (ICC) is the metric that was most used to investigate the reliability of individual differences. Magnetic resonance imaging (MRI) investigate the structural and functional changes of the brain in a non-invasive manner and has become the workhorse of neuroscience and psychology. There are already a number of studies that systematically investigated the test-retest reliability of structural and functional MRl metrics but seldom has consider whether the scanner, scanning protocol will affect the investigation of individual differences. In order to give a preliminary answer to this question, this study first established a multi一sites multi-sessions MRI dataset and taking the macro-scale functional networks of human brain as templates and create a multi-layers linear mixed effects modeling on structural morphology and intrinsic functional metrics, quantitative research on inter-individual differences, intra-individual differences, site differences and their influence on different reliability.

Study one introduces the design of a multi一sties mulit-sessions dataset that we collected in last three that is called R3BRAIN (reliability, reproducibility and replicability) and built a linear mixed effect model based on this dataset. In this study 51 participants who finished all 4 times 3T scans were included. Individual differences of structural morphology metrics were investigated in this study. Results showed that individual differences were affected by site factors, cortical thickness suffered more than cortical area on large-scale networks. Inter-sessions reliability were almost perfect across all networks (ICC>0.8) mean while inter-sites reliability were all above 0.6.

In study two we investigate the amplitude, inter-network connectivity, intra-network connectivity derived by implement dual-regression on Yeo's 17 networks. Results showed that amplitude and inter-networks connectivity have higher individual differences after regress global signal and showed less influences by site factors. Intra-network connectivity showed less influence by site factors. Brain networks of control and default have higher individual differences. Area where shows ICC above 0.4 become smaller when introduces site factors.

In study three the sex differences of large-scale network morphology and function metrics was examined by a linear mixed model. The results show that gender-related effects are drives by the global characteristics, and after such global differences are properly controlled, the differences in large-scale networks between men and women are no longer significant. More specifically, the morphological differences between men and women are mainly due to males have larger brain volume. That differences were mainly affected by the difference in surface area. Sex differences are also manifested in the functional brain network level where males have stronger inter-network connectivity between the control network, the attention network and the somatosensory movement network. Intra-networks differences mainly showed on somatosensory network in which male showed stronger connectivity/ The above-mentioned gender differences are no longer significant after controlling the global characteristics accordingly, and are reproducible at different sites.

To sum up, this study provides a reference for multi-site research. It is found that the individual difference measurement of brain morphology has almost perfect test-retest reliability, while the measurement of functional individual difference shows moderate to substantial test-retest reliability. The signal-to-noise ratio (SNR) is an important factor affecting the test-retest reliability. The individual difference measurement of morphology metrics has considerable and near-perfect inter-site test-retest reliability. Functional individual difference measurement's site reliability shows network-dependent differences, control network, dorsal attention network, default network have considerable higher reliability. SNR is also an important factor that affects site test-rest reliability. The cortical morphology of the parietal memory network has high individual differences. The associative cortical network has higher inter-individual differences in the spontaneous functional signal than the primary cortical network, while the brain network gender differences are mainly driven by the global metrics of structure and function.