中国科学院自动化研究所机构知识库

       人类复杂并且多样的行为表现依托于人脑复杂的功能，而人脑之所以能执行更为复杂的功能，这与人脑特有的结构基础是分不开的。人类的行为究竟如何依托于脑的功能，而脑的功能又是如何起源于脑的结构基础，这些问题仍然是当今神经科学领域中的谜题。本研究也将围绕“脑结构—―脑功能——人类行为”这条主线，试图去探索这三者之间的联系，从而为认识脑功能和人类行为的起源提供新的见解。之前的研究表明脑结构连接和视觉功能存在着密切的关系，但是在功能灵活性更强并且个体差异更大的联络皮层上，这种密切关系是否存在仍然是未知的。除了高级联络皮层所表现出的更大的结构和功能个体差异，人类行为也在多种领域表现出了很大的个体差异。研究者们对这些广泛观察到的个体差异的起源和意义仍然没有完全理解，并且通常使用组平均方法忽略掉这些个体差异。因此理解人类行为个体差异的脑结构和功能起源也十分重要。为了回答以上问题，本文主要进行了如下工作：

       我们利用弥散和功能磁共振成像探索并验证各个脑区的结构连接特征在多大程度上预测各个脑区的多种功能。通过使用线性预测模型，我们在来自HCP数据集的七种不同功能领域上，检验了整个脑皮层上是否普遍存在脑结构连接和功能激活之间的关系。我们发现了脑结构连接在多种功能领域上对脑功能激活的预测能力都超过随机对照模型，但是，在功能个体差异较大的联络皮层上，脑结构连接的预测能力要低于其在功能个体差异较小的感觉运动皮层上的预测能力。我们的结果还表明这种脑结构连接和功能激活关系的层级结构还与多种其他的脑组织结构有关，比如髓鞘化，功能灵活性和功能个体差异等。当前的这些发现系统地描述了脑结构连接和功能激活之间的关系，并且为理解脑的结构和功能组织提供了重要的见解。

       之前的关于脑结构和人类行为关系的研究常常采用基于小样本的相关性分析，这种研究方式对新的样本没有泛化性能的保证并且常常导致过高的估计。我们主要在预测模型框架下探索人类行为中的个体差异在多大程度上与脑结构连接中的个体差异相关，并且进一步发现多种行为背后的脑结构连接基础。我们使用了来自HCP数据集的1009名被试的行为和弥散成像数据，并且通过偏最小二乘回归模型揭示出了不同行为背后的脑结构连接基础，我们进一步证实了通过预测模型提取出的脑结构连接模式可以被用于对未知被试的多种行为进行预测。我们的结果系统地描述了脑结构连接和多种行为之间的关系，并且提供了一种将人类行为和脑解剖结构联系起来的途径。

       基于人类的行为是受到脑功能的影响这个基本假说，我们探索了脑功能激活中的个体差异在多大程度上关联了多种领域的行为。通过使用偏最小二乘回归模型，我们发现了七种任务状态下脑功能激活的个体差异模式与五大领域的多种行为都存在相关性，并且我们证实了这些脑功能激活个体差异模式可被用于预测未知被试的行为。我们还揭示了这些与行为相关的脑功能激活个体差异模式可以在不同任务状态下进行迁移，并且可以被用于重建个体脑功能激活。重建的个体脑功能激活保留了在组平均中所损失的一些个体差异，并且可以被用于个体功能定位。这些结果详细描述了脑功能激活中的个体差异和行为之间的关系，并且在个体化的精准医疗中显示出了潜在的应用价值。最后我们通过将精神分裂症患者受损的工作记忆功能和疾病行为表现之间建立联系，揭示了精神分裂症患者拥有的三网络工作记忆损伤模式，这也为理解精神分裂症的发病机制提供了重要见解。

       The complex and flexible behaviors of human rely on the complex brain functions, and the merit that human brain can execute complex functions should be attributed to the structural organizations of human brain. However, the extent to which that human behaviors rely on the brain functions and how the brain structural substrates give rise up to the brain functions still remain to be the mysteries of modern neuroscience. This work will focus on the axis linking brain structure, brain function and human behavior to explore the relationship between them, thus providing new insights into the origins of brain function and human behavior. Previous studies discovered a close relationship between anatomical connectivity and visual functions. However, whether this close connectivity-function relationship can generalize to association cortex that is both functionally flexible and variable is still unknown. In addition to the anatomical and functional variability in high order association cortex, human behaviors also exhibited great variability in various domains. However, the origin and meaning of these individual differences are not fully understood and are often neglected through the group averaging. Understanding the brain origins of the inter-subject variability in human behaviors is thus very important. To address the above questions, this work mainly concentrates on the following aspects:

       We validated the extent to which each region’s anatomical connectivity can predict that region’s various functions in a vertex-wise manner via diffusion and functional magnetic resonance imaging. Using a linear prediction model, we examined whether the connectivity-function relationship is universal in the whole cortex across seven functional domains from the Human Connectome Project dataset. We found that anatomical connectivity possessed predictive ability that was statistically better than control models across various cognitive domains, but, the predictive ability of anatomical connectivity was lower in the association cortex that had a large inter-subject task variation than in the sensory-motor cortex that had a small inter-subject task variation. Our results also revealed that the hierarchy of cortical areas in the connectivity-function relationship was correlated to other brain organizations such as myelin map, functional flexibility and functional variability. The current findings delineate the first comprehensive picture of the relationships between functional activations and anatomical connectivity profile across the whole cerebral cortex, and provide important insights into the understanding of anatomical and functional organization of the human brain.

       Previous works that investigated the relationship between brain structure and human behavior often adopted a small number of subjects in a correlational analysis, which is not generalizable to new subjects and leads to overestimation. We aim to explore the extent to which the variability in human behaviors is related to the variability in anatomical connectivity under a framework of prediction and identify the structural substrates of individual differences in multiple behaviors. Using a large behavioral and diffusion imaging data of 1009 subjects from the Human Connectome Project, we revealed patterns of anatomical connectivity that underlay different behaviors via a model of partial least square regression, and demonstrated that these patterns of anatomical connectivity extracted from the prediction model could be used to predict multiple behavioral traits of unseen subjects. Our results describe a comprehensive relationship between multiple behaviors and anatomical connectivity, and provide a way of linking human behavior to brain anatomy.

       Based on the fundamental assumption that human behavior is determined by the underlying brain functions, we investigated the extent to which that these individual differences in brain activity are related to a wide domain of behavioral traits. Using the model of partial least square regression, we discovered that the patterns of individual differences in brain activity induced at seven task state were related to multiple behavioral traits from five domains, and we demonstrated that these patterns of individual differences could be used to predict behavioral traits of unseen subjects. We also revealed that the behavior-relevant individual differences transferred between different task states and could be used to reconstruct the individual brain activity. The reconstructed individual brain activity kept some individual differences that were lost in the group average and could serve as an individual functional localizer. The results delineate a comprehensive relationship between individual differences in brain activity and behavioral traits, and show potential use in precise, personalized medicine. Finally, by linking the dysfunction of working memory of schizophrenia with the illness behaviors of schizophrenia, we revealed a triple network dysfunction of working memory in schizophrenia patients, which also provides important insights into the understanding of the mechanism of schizophrenia.