用100万张脑组织切片训练模型，实现大脑皮层细胞级结构的智能分析。

Studying the cellular architecture of the human cerebral cortex is essential for understanding how the brain is organized from the micro to the macro level, and how it functions. However, investigating complex texture patterns in histological images using automatic methods that can be scaled across whole brains remains a challenge. Here we introduce CytoNet, a foundation model trained on 1 million unlabeled microscopic image patches from over 4,000 histological sections from nine postmortem brains, and evaluated on over 2,000 sections from five additional brains excluded from self-supervised pretraining. Using co-localization in the cortical sheet for self-supervision, CytoNet learns to encode complex cellular patterns into expressive and anatomically meaningful feature representations. CytoNet supports multiple downstream applications, including area classification, laminar segmentation, quantification of microarchitectural variation, and exploratory mapping of cortical subdivisions. Functional parcellation analyses provided parcellation-dependent evidence for links between cytoarchitecture and macroscale functional organization. Together, these results establish CytoNet as a unified framework for scalable analysis of cortical microarchitecture and for testing links between cellular architecture and structure-function organization in the human cerebral cortex.