Duda Marlena, Chen Jiayu, Belger Aysenil, Ford Judith, Mathalon Daniel, Preda Adrian, Turner Jessica, Van Erp Theo, Pearlson Godfrey, Calhoun Vince D
Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia, USA.
Department of Psychiatry, University of North Carolina, Chapel Hill, North Carolina, USA.
Hum Brain Mapp. 2025 Aug 1;46(11):e70302. doi: 10.1002/hbm.70302.
The precise relationship between brain structure and function has been investigated through a multitude of lenses, but one detail that is held constant across most neuroimaging studies in this space is the identification of a singular structural basis set of the brain, upon which functional activation signals can be reconstructed to examine the linkage between structure and function. Such basis sets can be considered "functionally independent", as they are derived through structural data alone and have no explicit association to functional data. Recent work in multimodal fusion has facilitated a more integrated view of structure-function linkages by enabling the equal contribution of both modalities to the joint decomposition, resulting in components that are independent within modality but co-vary closely across modalities. These existing symmetric fusion approaches thus identify structural bases given an associated functional context. In this work, we consider an additional layer of precision to the investigation of structure-function coupling by studying these context-dependent linkages in a time-resolved manner. In other words, we ask which features of brain structure become (or remain) salient given the dynamically changing functional contexts (i.e., dynamic functional connectivity states, task structure, etc.) the brain may pass through during a given fMRI scan. We introduce "dynamic fusion", an ICA-based symmetric fusion approach that enables flexible, time-resolved linkages between brain structure and dynamic brain function. We show evidence that temporally resolved and functionally contextualized structural basis sets can accurately reflect dynamic functional processes and capture diagnostically relevant structure-functional coupling while detecting nuanced functionally driven structural components that cannot be captured with traditionally computed structural bases. Lastly, differential analysis of component stability across repeated scans from a control cohort reveals that the organization of static and dynamic structure-function coupling falls along unimodal/transmodal hierarchical lines.
人们已经通过多种视角研究了大脑结构与功能之间的精确关系，但在这个领域的大多数神经成像研究中，一个保持不变的细节是确定大脑的单一结构基础集，在这个基础集上可以重建功能激活信号，以检验结构与功能之间的联系。这样的基础集可以被认为是“功能独立的”，因为它们仅从结构数据中得出，与功能数据没有明确关联。多模态融合方面的最新工作通过使两种模态在联合分解中做出同等贡献，促进了对结构 - 功能联系的更综合的看法，从而产生了在模态内独立但跨模态紧密协变的成分。因此，这些现有的对称融合方法在给定相关功能背景的情况下识别结构基础。在这项工作中，我们通过以时间分辨的方式研究这些上下文相关的联系，为结构 - 功能耦合的研究增加了一层额外的精确性。换句话说，我们要问的是，在给定大脑在一次功能磁共振成像扫描中可能经历的动态变化的功能背景（即动态功能连接状态、任务结构等）下，大脑结构的哪些特征会变得（或保持）显著。我们引入了“动态融合”，这是一种基于独立成分分析（ICA）的对称融合方法，它能够在大脑结构和动态脑功能之间实现灵活的、时间分辨的联系。我们证明，时间分辨且功能上下文相关的结构基础集能够准确反映动态功能过程，并捕捉具有诊断相关性的结构 - 功能耦合，同时检测到传统计算的结构基础无法捕捉的细微功能驱动的结构成分。最后，对来自对照组的重复扫描中成分稳定性的差异分析表明，静态和动态结构 - 功能耦合的组织沿着单模态/跨模态层次线分布。
