虚拟数字脑软件平台的研究与开发

目的:为脑科学和类脑人工智能提供一个软件研究平台。方法:首先，基于预处理后的静息态功能磁共振数据和因果连接法构建因果连接网，利用最小均方误差算法获得节点神经活动预测模型中的回归系数。然后，基于因果连接网和预测模型构建虚拟数字脑。最后，在VC++12.0开发环境下，利用C/C++编程语言开发出能实现虚拟数字脑各项功能的软件包。结果:同欧洲的虚拟大脑软件平台相比，该软件包操作简单，可用于脑科学和类脑人工智能的研究。结论:初步的实践结果证明了该软件在脑科学研究中的有效性和实际应用价值。

Research and development of virtual digital brain software package

Abstract: Objective To provide a software research platform for brain science and brain-inspired artificial intelligence. Methods Firstly, a causal connection network was constructed on the basis of preprocessed resting-state functional magnetic resonance imaging data by a causal connectivity method. The regression coefficients of a nodal neural activity prediction model were obtained by the least-mean-square error algorithm. Then, the virtual digital brain was constructed based on the causal connection network and the prediction model. Finally, a software package which can implement all functions of virtual digital brain was developed by C/C++ programming language in the VC++12.0 development environment. Results Compared with the European virtual brain software platform, the proposed software package was easy to operate and could be applied to the researches on brain science and brain-inspired artificial intelligence. Conclusion The results of preliminary practices confirm the validity and application value of the software in the research on brain science.