Desai Neel, Baladandayuthapani Veera, Shinohara Russell T, Morris Jeffrey S
Division of Biostatistics, University of Pennsylvania, 423 Guardian Drive, Philadelphia, PA, 19104, United States.
Department of Biostatistics, University of Michigan Ann-Arbor, 1415 Washington Heights, Ann Arbor, MI 48109, United States.
Biostatistics. 2024 Dec 31;26(1). doi: 10.1093/biostatistics/kxaf002.
Assessing how brain functional connectivity networks vary across individuals promises to uncover important scientific questions such as patterns of healthy brain aging through the lifespan or dysconnectivity associated with disease. In this article, we introduce a general regression framework, Connectivity Regression (ConnReg), for regressing subject-specific functional connectivity networks on covariates while accounting for within-network inter-edge dependence. ConnReg utilizes a multivariate generalization of Fisher's transformation to project network objects into an alternative space where Gaussian assumptions are justified and positive semidefinite constraints are automatically satisfied. Penalized multivariate regression is fit in the transformed space to simultaneously induce sparsity in regression coefficients and in covariance elements, which capture within network inter-edge dependence. We use permutation tests to perform multiplicity-adjusted inference to identify covariates associated with connectivity, and stability selection scores to identify network edges that vary with selected covariates. Simulation studies validate the inferential properties of our proposed method and demonstrate how estimating and accounting for within-network inter-edge dependence leads to more efficient estimation, more powerful inference, and more accurate selection of covariate-dependent network edges. We apply ConnReg to the Human Connectome Project Young Adult study, revealing insights into how connectivity varies with language processing covariates and structural brain features.
评估大脑功能连接网络如何因人而异，有望揭示重要的科学问题，比如贯穿一生的健康大脑衰老模式，或与疾病相关的连接障碍。在本文中，我们介绍了一种通用回归框架——连接性回归（ConnReg），用于在考虑网络内部边间依赖性的同时，将特定个体的功能连接网络对协变量进行回归分析。ConnReg利用Fisher变换的多元推广，将网络对象投影到一个替代空间，在该空间中高斯假设成立且自动满足半正定约束。在变换后的空间中进行惩罚多元回归，以同时在回归系数和协方差元素中引入稀疏性，协方差元素捕获网络内部的边间依赖性。我们使用置换检验进行多重性调整推断，以识别与连接性相关的协变量，并使用稳定性选择分数来识别随选定协变量变化的网络边。模拟研究验证了我们提出方法的推断特性，并展示了估计和考虑网络内部边间依赖性如何导致更有效的估计、更强大的推断以及更准确地选择与协变量相关的网络边。我们将ConnReg应用于人类连接组计划青年成人研究，揭示了连接性如何随语言处理协变量和大脑结构特征而变化的见解。
