中国现代神经疾病杂志 ›› 2025, Vol. 25 ›› Issue (2): 112-120. doi: 10.3969/j.issn.1672-6731.2025.02.003

• 数智神经外科学 • 上一篇    下一篇

2 超算脑模拟技术应用进展

孙哲()   

  1. 1138421 东京, 日本顺天堂大学医学研究科数据科学系计算生物工程研究室

Progress on the application of supercomputer brain simulation technology

Zhe SUN()   

  1. Computational Bioengineering Laboratory, Faculty of Health Data Science, Graduate School of Medicine, Juntendo University, Tokyo 1138421, Japan
  • Received:2025-02-15 Online:2025-02-25 Published:2025-04-07

摘要:

通过高性能计算平台进行大规模脑模拟研究成为神经科学与计算科学交叉领域的重要趋势。伴随新一代高性能计算平台的崛起以及脑科学与类脑研究的推进,在多尺度、跨模态数据的支持下,构建更接近生物真实性的全脑或局部回路仿真模型,为阐明脑功能机制、揭示脑疾病发病机制、助力类脑智能技术发展带来革新性机遇。然而如何在巨大的计算负载、繁杂的多模态数据管理及跨学科协作中持续取得突破,仍存在诸多挑战。本文综述大规模脑模拟的理论基础与关键技术、常用的大规模脑模拟平台与软件工具、可使用的脑模拟数据和资源、脑疾病的超算脑模拟研究、类脑智能技术与脉冲神经网络训练在高性能计算平台的应用,讨论面临难题与潜在解决方案并展望未来发展方向,认为E级超算与多模态大数据融合将为全面理解仿真大脑提供前所未有的契机,也为个性化医疗与新一代人工智能注入持续的动力。

关键词: 脑疾病, 计算机模拟, 人工智能, 综述

Abstract:

High performance computing (HPC) is transforming the field of large-scale brain simulation by enabling the integration of multi-scale computational modeling with massive neuroscience data. With advanced HPC resources, researchers can simulate neural activities from ion-channel dynamics to whole-brain network interactions, thereby illuminating the mechanisms underlying cognition, neural disorders, and emerging neuromorphic intelligence. This review examines the theoretical principles and technical foundations of supercomputer brain simulation, including distributed parallel algorithms, graphics processing unit (GPU)-based acceleration, and multimodal data management. It also surveys prominent simulation platforms such as NEST, NEURON, and The Virtual Brain (TVB), highlighting their strengths in modeling spiking neuronal network (SNN), multicompartmental neurons, and large-scale functional connectivity, respectively. Furthermore, we discuss the practical applications of these simulations in elucidating disease mechanisms in Alzheimer's disease (AD), Parkinson's disease (PD), autism spectrum disorder (ASD), schizophrenia, and epilepsy. Special emphasis is placed on how supercomputer brain simulation assists in virtual drug screening, optimizing deep brain stimulation parameters, and supporting digital twin approaches for personalized medicine. Finally, we address the critical challenges and future directions in this rapidly evolving domain, including the trade-off between computational cost and biological realism, data integration and validation, and the necessity for interdisciplinary collaboration. The advent of exascale supercomputers and the convergence of neuroinformatics and machine learning (ML) are poised to propel brain simulation research toward unprecedented clinical and scientific breakthroughs.

Key words: Brain diseases, Computer simulation, Artificial intelligence, Review