Dynamic Analysis of Neuronal Pathway by Engineering Identification Method for Non-linear Signal Processing

利用信息学的工程方法，将神经元通路分解为若干个单元，并揭示其内在机制，首先对夹层系统--两个线性滤波单元级联夹着一个非记忆非线性单元--的分解精度进行了评估。对于第一和第二Wiener核的z变换方法，确定了用于部分感觉系统的元素估计所需的时间域数据的长度。其次，在去掉Wiener核函数的统计响应特性后，提取昆虫风受体脉冲密度编码机制的动力学细节。脉冲密度编码滞后的起始期是脉冲密度编码滞后的原因。第三，研究了蟋蟀颈区感觉中间神经元的传入通路。结果发现，尖峰和非尖峰中间神经元具有不同的传入通路组织。第四，提出了一种三明治系统分解的新方法，即利用Laguerre函数对Wiener核进行扩展。第五，采用单电极电流钳记录方法，测定了蟋蟀颈髓感觉中间神经元突触电流向棘串序列的信号传递。突触后膜被确认为整合的无源低通滤波器，而棘波起始区被确认为带通有源滤波器。第六，通过对Hodgkin-Huxley模型的数值求解，研究了随机共振和噪声对脉冲编码的影响。噪声对尖峰信号的编码起着至关重要的作用，而不是恶化信号。

Neuronal pathways, which operate as non-linear signal processors, were decomposed into several elements by using engineering method od informatics, and the internal mechanisms were revealed.First, accuracy of decomposition into sandwich system, a cascade of two linear filter elements sandwiching a nonmemory nonlinear element, was evaluated. The length of time domain data necessary for the element estimation ofinsect sensory system was determined for the z-transform method of the first and second Wiener kernels. Second, details of dynamics of the pulse density encoding mechanism of the insect wind receptor were extracted after removing the statistical response properties given by Wiener kernels. The refractory period of spike initiation was the origin of the hysteresis in the pulse density incoding.Third, input pathways to cercal sensory interneurons of cricket were studied. It was found that the spiking and nonspiking interneurons have different organizations of input pathway. Fourth, a new method for sandwich system decomposition, in which Laguerre functions were used to expand Wiener kernels, was developed. Fifth, by using the single electrode current clamp recording method, signal transfer from the synaptic current to spike train sequence in the cricket cercal sensory interneuron was determined. The post-synaptic membrane was confirmed as an integrating, passive low-pass filter, while the spike initiation area was identified as a band-pass active filter. Sixth, the stochastic resonance and the effects of noise in spike encoding were studied by numerical solution of the Hodgkin-Huxley model. Noise plays an essential role for spike encoding rather than deteriorating the signal.