Understanding how one brain region exerts influence over another in vivo is profoundly constrained by models used to infer or predict directed connectivity. Although such neural interactions rely on the anatomy of the brain, it remains unclear whether, at the macroscale, structural (or anatomical) connectivity provides useful constraints on models of directed connectivity. Here, we review the current state of research on this question, highlighting a key distinction between inference-based effective connectivity and prediction-based directed functional connectivity. We explore the methods via which structural connectivity has been integrated into directed connectivity models: through prior distributions, fixed parameters in state-space models and inputs to structure learning algorithms. Although the evidence suggests that integrating structural connectivity substantially improves directed connectivity models, assessments of reliability and out-of-sample validity are lacking. We conclude this Review with a strategy for future research that addresses current challenges and identifies opportunities for advancing the integration of structural and directed connectivity to ultimately improve understanding of the brain in health and disease.

了解一个大脑区域如何在体内对另一个大脑区域产生影响，受到用于推断或预测定向连接的模型的严重限制。尽管这种神经交互依赖于大脑的解剖结构，但目前尚不清楚在宏观尺度上，结构（或解剖学）连接是否为定向连接模型提供了有用的约束。在这里，我们回顾了这个问题的研究现状，强调了基于推理的有效连接和基于预测的定向功能连接之间的关键区别。我们探索了将结构连接集成到有向连接模型中的方法：通过先验分布、状态空间模型中的固定参数以及结构学习算法的输入。尽管证据表明整合结构连接大大改善了定向连接模型，但缺乏对可靠性和样本外有效性的评估。我们以未来研究策略结束本综述，该策略应对当前挑战并确定推进结构和定向连接整合的机会，以最终提高对健康和疾病中大脑的理解。