基于同步多维数据流的脑网络动态特征辨识方法研究

Research of dynamic characteristic identification method for human brain network based on multidimensional synchronization data flow

针对人脑实时变化的特性,为了更好地观测和描述人脑网络的动态特征,在基于功能磁共振成像的脑功能网络重构技术基础上,给出了一种人脑网络动态特征辨识方法。首先利用同步多维数据流的即时更新能力,将在静息态功能磁共振成像数据采集区间上的血氧水平依赖信号由大时间序列分解重构为每个采样点上的小时间窗口序列,构建连续时间点上的状态观测窗口,从而实现对人脑功能共振信号的特定时间状态辨识;然后运用相关分析对状态观测窗口信号进行分析,得到单状态观测矩阵,最终构建全脑在整个数据采集区间上的动态特征矩阵。实验结果显示该方法可以为人脑网络的动态特征观测和描述提供一种有效手段,也为进一步研究人脑网络的动态特征演变奠定了基础。

Focusing on the real-time change characteristic of the human brain, this paper proposed a dynamic characteristic identification method to extract and describe the dynamic properties of the human brain network based on functional magnetic resonance imaging technology. Firstly, the method used multidimensional synchronization data flow analysis technology with real-time updating capability to decompose the blood oxygen level-dependent signal sequences on the whole data acquisition time into small time windows sequences on each sample point. Then it got the continuous state observer windows which could realize the specific time point status extraction of the human brain functional magnetic resonance signal. Secondly, it used a correlation analysis technology to analyze the signals in adjacent state observer windows to obtain a single state observer matrix which could describe the state in a single sample point. Finally, it had the entire dynamic feature matrix during the data collection by combining multiple state observer matrix. The experiment results show this method provides an effective means for dynamic characteristic observation and description of human brain functional network and also gives foundations for further study for human brain network dynamic properties.