神经计算建立在不同神经元之间跨尺度的复杂相互作用之上。解码神经元计算规则，需要根据神经元的功能角色来识别神经元，并重建它们的微观连接。尽管大规模的神经元监测（包括电生理学和光学监测）为我们提供了新的机会，尚需要实时高通量计算和闭环神经调控来控制和标记特定功能类型的神经元。我们曾经借鉴天文学研究经验，发展了一套FPGA-GPU系统，实时解析每秒超过500 MB的全脑功能成像数据流，提取全脑近十万个神经元的活动，从而实现了闭环的精准光遗传学调控。在此基础上，我们正在实时追踪全脑尺度神经元集群构成的动态演化，并开发光控标记方法和基于集群动态的神经调控，试图理解神经元之间的突触连接和相互作用是如何产生全脑尺度功能状态的。

Neural computation is based on neuronal interaction across different spatial scales. To discriminate the functional roles of neurons and to reconstruct their ultra-structural connections, are required for the dissection of neural computation algorithms. While this has been permitted by large-scale neuronal monitoring, both electrophysiology and optical imaging, real-time processing of the recorded data of high-throughput and closed-loop modulation are required to interrogate the neurons of specific functional roles. Based on previous knowledge in astronomy detection, we developed an FPGA-GPU system and decoded brain-wide neuronal imaging at 500 MB/s in real time. This enabled closed-loop patterned optogenetic manipulation tuned to the activity fluctuation of the hundreds of thousands of neurons across the zebrafish brain. Furthermore, we are tracking the dynamic evolution of brain-wide neuronal ensembles. We are also working on optical ultrastructural tagging methods for understanding the synaptic connection basis underlying brain-wide functional dynamics.