从儿童到青少年时期，脑白质作为神经环路的解剖基础经历广泛而剧烈的生物物理变化，如髓鞘形成、突触修剪、轴突直径及密度增加，以促进脑区间快速的神经信号交流和功能同步化，支持多种认知功能和行为能力的快速发展。基于弥散磁共振成像（diffusion magnetic resonance imaging，dMRI）技术，可以非侵入地刻画宏观尺度下的人脑白质结构连接组，进而系统揭示大脑神经环路的发育机制和认知功能的结构基础。尽管该方法在认知神经科学领域备受关注，但对人脑白质连接组的发育规律及其与认知发展之间的关系仍不清楚。本论文将针对儿童青少年时期人脑白质连接组的发育轨迹、结构与功能网络耦合模式、以及基于脑白质网络的个体化认知预测三个方面开展系统研究。
人脑白质结构连接组的纵向发育轨迹。在儿童到青少年时期，人脑白质结构连接组的多尺度发育重塑及其潜在的转录组和细胞机制是未知问题。本节研究利用儿童学习能力与脑发育项目（Children School Functions and Brain Development Project，CBD）数据集中604名健康被试（6~13岁）三个时间点的纵向dMRI数据，系统绘制了人脑白质连接组的全局、局部和连接水平的发育轨迹，并研究了其与转录组和细胞结构的关系。研究发现，大多数连接组属性遵循线性发育轨迹。在区域水平上，不同脑区节点属性的发育速率存在异质性，发育速率较高的区域主要位于枕叶皮质、梭状回、颞上回、扣带回、海马和楔前叶。通过连接组-转录组分析，揭示了人脑白质连接组的空间发育模式可能受到转录组结构的调控，正相关基因参与离子运输和发育等相关通路，并在兴奋性和抑制性神经元中表达；而负相关基因富集于突触和发育相关通路，并在星形胶质细胞、抑制性神经元和小胶质细胞中表达。此外，这种发育模式也与髓鞘化程度和特定皮质分层的厚度有关，表明在细胞水平上存在潜在的微观机制。最后，在人类连接组发育项目（Lifespan Human Connectome Project in Development，HCP-D）的179名健康被试（6~13岁）的横断面dMRI数据中验证了这些发现。总的来说，通过绘制纵向队列中人脑白质连接组的发育轨迹，并将其与转录组和细胞结构图谱联系起来，本研究提供了从童年到青春期宏观白质连接组的发育规律及潜在遗传学和细胞层面解释。
人脑结构与功能连接组发育的时空耦合模式。结构与功能耦合可以描述结构环路如何塑造跨皮层神经活动的大规模功能组织。然而，人脑连接组的结构与功能耦合在发育过程中如何发展及其与认知差异和转录组结构的关系尚不清楚。本节研究使用来自HCP-D数据集中439名健康被试（5.67~21.92岁）的多模态磁共振成像数据，通过结合皮层内和皮层外结构连接组来预测功能连接组表征结构与功能耦合。研究结果显示，结构与功能耦合度在视觉和感觉运动网络与其它网络相比更强，这与进化扩张、髓鞘化程度和功能主梯度一致。随着大脑不断发育，结构与功能耦合表现出以正增加为主的空间异质性改变，广泛分布于感觉运动网络、额顶网络、背侧注意网络和默认网络。此外，结构与功能耦合可以显著预测总体认知的个体差异，其中额顶网络和默认网络编码更高的权重。最后，结构与功能耦合的异质发育与在少突胶质细胞相关通路中富集的基因正相关，而与星形胶质细胞相关基因负相关。本节研究进一步提供了对结构与功能耦合发育原理的见解。
基于人脑白质结构连接组的个体认知预测。基于机器学习方法，人脑白质连接组特征（如全局、局部、连接属性）可作为预测年龄和认知能力等个体化测度的影像学标记物。然而，基于连接组的预测模型本质上网络构建方法和回归算法选择的影响，缺乏针对人脑白质连接组的预测框架的系统评估。基于北京老年脑健康计划（Beijing Aging Brain Rejuvenation Initiative，BABRI）和人脑连接组老化项目（Lifespan Human Connectome Projects in Aging，HCP-A）的两个独立数据集（BABRI：633名认知正常被试；HCP-A：560名认知正常被试），考虑了两种节点定义策略和七种连接定义策略分别构建网络，基于八种回归算法对年龄和四种认知功能进行预测。研究发现人脑白质连接组对个体年龄和认知功能（特别是执行功能和注意功能）有着较好的预测能力。在网络构建层面，不同的节点定义策略在预测性能上存在显著差异，不同采集参数的dMRI对纤维重建方法和连接加权策略有各自偏好。在回归算法层面，MLP和Elastic-Net算法有更加准确和稳健的预测性能。根据以上方法学评价结果，利用HCP-D发育数据集（439名健康被试）的白质结构连接组特征对广泛的认知评分进行预测评估，结果表明白质连接组可以显著预测发育群体的不同认知测量，且对不同测量的预测贡献存在高度一致性。综上所述，本节研究确定了人脑白质网络构建和机器学习回归算法对预测性能的影响，同时揭示了人脑白质连接组在发育群体中对不同认知的预测效力，为后续的人脑白质连接组和认知预测相关研究提供了重要的方法学基础。
本论文研究了儿童到青少年时期的人脑结构白质连接组的发育轨迹、人脑结构与功能连接组的时空耦合模式和基于脑白质结构连接组的个体认知预测，为更好地理解大脑发育补充了一个新视角，也为脑白质结构连接组的网络构建和认知预测提供了方法学参考。

From childhood to adolescence, the white matter in the brain undergoes extensive and dramatic biophysical changes, such as myelination, synaptic pruning, an increase in axon diameter and density, to facilitate rapid neural signal communication and functional synchronization between brain regions, supporting the rapid development of various cognitive functions and behavioral abilities. Using diffusion magnetic resonance imaging (dMRI) technology, it is possible to non-invasively characterize the macro-scale human brain white matter structural connectome, thereby systematically revealing the developmental mechanisms of brain neural circuits and the structural basis of cognitive functions. Despite the considerable attention this method has received in the field of cognitive neuroscience, the developmental patterns of the human brain white matter connectome and its relationship with cognitive development remain unclear. This paper will systematically study three aspects related to the development trajectory of the white matter connectome in children and adolescents, the structural and functional connectome coupling patterns, and individualized cognitive predictions based on the white matter connectome.
Longitudinal developmental trajectory of human brain white matter structural connectome. During childhood to adolescence, the multiscale developmental reorganization of human brain WM connectome and its underlying transcriptional and cellular mechanisms are still unknown. In this study, longitudinal dMRI data from the Children School Functions and Brain Development Project (CBD) dataset were used, including 604 healthy subjects (aged 6 to 13), to systematically depict the developmental trajectories of global, local, and connection-level properties of human brain WM connectome. The relationship between WM connectome and transcriptional profiles and cellular structures was also investigated. The study found that most properties of the connectome followed linear developmental trajectories. At the regional level, there was heterogeneity in the developmental rates of different brain regional nodal properties, with higher developmental rates observed in regions primarily located in the occipital cortex, cingulate gyrus, superior temporal gyrus, precuneus, hippocampus, and precentral gyrus. Through connectome-transcriptome analysis, the spatial developmental patterns of WM nodal efficiency were found to be potentially regulated by transcriptional structures. Positively correlated genes were involved in ion transport, development-related pathways, and expressed in excitatory and inhibitory neurons, while negatively correlated genes were enriched in synaptic and developmental-related pathways and expressed in astrocytes, inhibitory neurons, and oligodendrocytes. Furthermore, these developmental patterns were related to myelin content and thickness of specific cortical laminas, suggesting potential microscale mechanisms at the cellular level. Finally, these findings were validated in cross-sectional dMRI data from 179 healthy subjects (aged 6 to 13) from the Lifespan Human Connectome Project in Development (HCP-D) dataset. In conclusion, by delineating the developmental trajectories of human brain WM connectome in a longitudinal cohort and linking them to transcriptional profiles and cellular structural maps, this study provides insights into the potential genetic and neural mechanisms underlying macroscopic WM connectome development from childhood to adolescence.
Spatiotemporal coupling patterns of the human brain structural and functional connectome development. Structure-function coupling refers to how structural circuits shape large-scale functional organization of neural activity across cortical areas. However, the developmental trajectory of structure-function coupling in the human brain connectome and its relationship with cognitive differences and transcriptional structures remain unclear. In this study, multimodal MRI data of 439 healthy participants (aged 5.67 to 21.92) from HCP-D were utilized to predict functional connectome based on the combination of intra-cortical and inter-cortical structural connectomes, characterizing the structure-function coupling. The results showed that the degree of structure-function coupling was stronger in visual and somatomotor networks compared to other networks, consistent with evolutionary expansion, myelin content, and functional gradients. With development, the spatial heterogeneity of structure-function coupling exhibited predominantly positive increases and was widely distributed in somatomotor networks, frontoparietal networks, dorsal attention networks, and default mode networks. Furthermore, structure-function coupling significantly predicted individual differences in total cognition, with higher weights encoded by frontoparietal networks and default mode networks. Finally, the heterogeneous developmental patterns of structure-function coupling were positively associated with genes enriched in oligodendrocyte-related pathways, while negatively associated with genes expressed in astrocytes. This study provides further insights into the developmental principles of structure-function coupling.
Individual cognitive prediction based on white matter structural connectome. Based on machine learning methods, features of the human brain WM connectome (such as global, local, and connection attributes) can serve as imaging biomarkers for personalized measurements, such as predicting age and cognitive abilities. However, the predictive models based on connectome are fundamentally influenced by the network construction methods and regression algorithms chosen, lacking systematic evaluations for prediction frameworks specifically targeting human brain WM connectome. Based on two independent datasets, the Beijing Aging Brain Rejuvenation Initiative (BABRI) and the Lifespan Human Connectome Projects in Aging (HCP-A) (BABRI: 633 healthy participants; HCP-A: 560 healthy participants), two node definition strategies and seven connection definition strategies were considered to construct networks. Eight regression algorithms were employed to predict age and four cognitive functions. The study found that human brain WM connectome exhibited good predictive ability for individual age and cognitive functions, particularly executive function and attention. At the network construction level, different node definition strategies showed significant differences in prediction performance, and dMRI with different acquisition parameters exhibited preferences for fiber reconstruction methods and connection weighting strategies. At the regression algorithm level, MLP and Elastic-Net algorithms demonstrated more accurate and robust prediction performance. Based on the results of the above methodological evaluation, the WM structural connectome features of the HCP-D dataset (439 healthy subjects) were used to predict a wide range of cognitive scores. The results showed that the WM connectome could significantly predict different cognitive measures in the developmental population, and the prediction contribution of different measures was highly consistent. In summary, this study identified the impact of human brain WM network construction and machine learning regression algorithms on prediction performance, providing an important methodological foundation for subsequent research on human brain WM conncetome and cognitive prediction.
This paper provides methodological references for the network construction of human brain WM structural connectome and cognitive prediction. It offers evidence for the developmental patterns, cognitive encoding abilities, and potential transcriptional and cellular structures of human brain WM connectome and structure-function coupling from childhood to adolescence. It also provides a new perspective for better understanding brain development.