一种基于混合模型的离子单通道信号重构算法研究

Algorithmic study on reconstruction of ionic single channel signal based on a novel hybrid model

细胞膜离子单通道信号是皮安级的随机离子电流,膜片钳技术记录中单通道电流往往被淹没在强背景噪声中。本文提出了一种隐马尔可夫模型(HMM)和径向基函数(RBF)神经网络相结合的混合模型,用于白噪声背景下细胞膜离子单通道电流的统计重构。该方法首先采用RBF网络强大的模式分类能力,确定离子单通道信号的电流幅值水平;然后利用HMM模型强时序信号处理能力,估计通道的动力学参数。在此基础上,从强噪声污染的膜片钳记录中统计重构理想化的通道电流信号。理论和仿真实验结果表明,在低信噪比情况下(SNR<5.0),该混合模型重构信号精度高,且具有较强的噪声鲁棒性。

signal of ionic single channel cell membrane is stochastic ionic currents on the order of lpA. Because of the weakness of the signal, the background noise always dominates by the patch-clamp recordings. A hybrid model of HMM and RBF neural network is presented to reconstruct the ionic single channel current under white background noise. In this method, the number of the channel current amplitude levels is determined by RBF neural network which has strong classifying ability. Then, the parameters are estimated utilizing the method to hidden Markov model processing dynamic time sequence. And the ideal channel current signal is reconstructed from patch-clamp recordings contaminated by noise. The theory and experiments have shown that the hybrid model performs effectively in the situation of low signal-to-noise ratio （SNR〈5.0） and has a high precision, strong noise robusticity.