摘要
在用事件相关电位研究与特定事件相关的脑活动时,由于实验过程中可能引入两个或多个事件(例如刺激和行为反应),导致叠加平均过程中真实的事件相关成份产生相互交叠。针对Zhang在1998年提出的两个事件相关成份分解算法中存在的病态问题,作者首先将维纳去卷积技术从一维线性系统的处理扩展到二维或多维线性系统的处理,其次应用维纳去卷积技术控制噪声成份对分解结果的影响,并推广维纳去卷积算法到三事件以及多事件相关成份分解。与传统的Tikhonov正则化方法比较,仿真实验结果表明,采用维纳去卷积技术能很好地恢复真实的仿真信号。最后将此算法应用于真实的实验数据,结果表明新算法得到的结果能更真实地反映大脑在处理相关事件时的脑活动过程。
Event-related potentials (ERPs) reflect brain activities related to specific behavioral events, and are obtained by averaging across many trial repetitions with individual trials aligned to the onset of a specific event, e.g., the onset of stimulus (s-lock) or the onset of behavioral response (r-lock). However, s-locked and r-locked ERP waveforms do not purely reflect, respectively, underlying stimulus (S-) or response (R-) component waveforms, because of their cross-contaminations in the recorded ERP waveforms. In 1998, Zhang proposed an algorithm to recover pure S-component waveform and pure R-component waveform from s- and r-locked ERP average waveforms. However, due to the inverse nature of this inverse problem, direct solution is sensitive to noise that can be shown to disproportionally affect low frequency components, hindering practical implementation of this algorithm. Here, the authors apply Wiener deconvolution to deal with noise in input data (two or multi-events), and obtained a stable solution that is robust against variances SNR comparing to Tikhonov regularization method. The method is demonstrated using three events experiment data. Result showed that the recovered event-related waveform based on Wiener deconvolution decomposition can reflect brain activity processing related event truly.
出处
《生物物理学报》
CAS
CSCD
北大核心
2010年第6期505-520,共16页
Acta Biophysica Sinica
基金
国家自然科学基金项目(60736029
60701015)~~