脳型リカレントニューラルネットワークの構造におけるマクロスケール機能的組織の発見

Understanding function from structure is a central topic in both neurobiology and artificial intelligence. In the human brain, macroscale functional organization, including functional parcellations, modules, and hierarchies, has been systematically linked to anatomical structure, providing an interpretable and clinically meaningful framework for linking structure to function. In contrast, although artificial neural networks are widely used as models of cognition, prior studies have not incorporated comprehensive macroscale brain structural constraints into model design, and artificial neural networks are rarely interpreted in terms of their structure. Consequently, it remains unclear whether principles linking structure and function in the human brain can be applied to artificial neural networks, and whether imposing brain-like structural constraints can give rise to findings of functional organization observed in the human cortex. Here, we introduce BrainRNN, a brain-like recurrent neural network architecture inspired by macroscale human cortical structure. We show that under structural constraints, BrainRNNs selectively regulate distribution of connectivity and increase activated association units for higher-order cognitive capacity. Moreover, we demonstrate structure-function coupling in BrainRNNs and show that macroscale functional organization, including functional modules and gradients, emerges along topographic and topological axes, closely mirroring empirical findings in the human cortex. Together, these results demonstrate how brain-like structure shapes functional organization and enables function to be discovered from the structure, highlighting the potential of structurally grounded artificial intelligence for neuroscientific research.