Research About Response Characteristics Of Local Field Potential Based On Information-preserving Empirical Mode Decomposition

Tuning characteristics of primary visual cortex (V1) to stimuli inside thereceptive field (RF) is of great significance to reveal neuronal signal processingmechanism. Neuronal signal recorded from V1is usually divided into local fieldpotential and spike, both of them have different characteristics in time and frequencydomain, but they both play important role in information spreading, at first spike isobject of research, and then local field potential gets more and more attention andnow both of them are two important kinds of neuronal signal. Research shows thatdifferent bands of LFP have different ability of representation stimuli. Because of thecomplexity of neuronal signal itself and lack of adaptivity of traditional method inprocessing non-stationary signal, tuning band separation of LFP has a variety ofresults. So it is meaningful to develop a method of frequency band selection that hasadaptivity.With an eye to the non-stationary of LFP, a method response signal extraction ofLFP using empirical mode decomposition and mutual information is developed,which can extract effective constituent from the complicated LFP. Then the extractedsignal is used to explore the spatial tuning and orientation tuning characteristics of V1,referring to spike from the same microelectrode array. At last, tuning characteristicsfrom analysis of gamma band by wavelet and second intrinsic mode function arecompared. We experiment on Long evens rats and record LFP and spike signal fromV1. The content of this paper includes the following two parts：Firstly, introduce the method of extraction of neural responding signals usinginformation-preserving empirical mode decomposition from local field potentials.Research shows that representation stimuli of different band in LFP is dynamic insome degree, which require a adaptive tool to process the signal. Empirical modedecomposition is adaptive in signal separation based on the amplitude of signal itself;mutual information analysis can indicate dependency of two data sets, regardless ofstation or nonstationary of the data, which is suitable to used to analysis LFP. LFP is first decomposed into several intrinsic mode functions(IMFs), then mutualinformation between LFP and each IMF and stimulus is calculated(recorded as MILFPand MIIMF), at last, adjusted mutual information between MILFPand each MIIMFiscalculated to facilitate the selection of the IMFs. Compared to traditional method,ours has the inherent advantage.Secondly, tuning characteristics analysis of V1based on the selected IMFs.Referring the spike from the same microelectrode array, spatial frequency tuning andorientation tuning of V1is explored, and comparison is carried out between resultsfrom gamma band by wavelet, second IMF. Results show that: stimuli relatedresponse signal extracted by our method is much prominent before and after stimuliboth in time domain and power spectrum; its consistency rate with MUA is72.9%inspatial frequency tuning and69.81%in orientation tuning; the selected IMFs retainthe90%information of the original signal, the average of tuning index (0.8423±0.039)is greater than IMF2(0.7951±0.043) and wavelet transform (0.6646±0.031), whichshows that extracting response band of LFP based on information-preserving EMDmethod is more advantage. This information-preserving EMD method can greatlysimplify the analysis of potentially complex signals and facilitate their interpretation.