Toward an Improvement of the Analysis of Neural Coding

Abstract

Machine learning and artificial intelligence have strong roots on principles of neural computation. Some examples are the structure of the first perceptron, inspired in the retina, neuroprosthetics based on ganglion cell recordings or Hopfield networks. In addition, machine learning provides a powerful set of tools to analyze neural data, which has already proved its efficacy in so distant fields of research as speech recognition, behavioral states classification, or LFP recordings. However, despite the huge technological advances in neural data reduction of dimensionality, pattern selection, and clustering during the last years, there has not been a proportional development of the analytical tools used for Time–Frequency (T–F) analysis in neuroscience. Bearing this in mind, we introduce the convenience of using non-linear, non-stationary tools, EMD algorithms in particular, for the transformation of the oscillatory neural data (EEG, EMG, spike oscillations…) into the T–F domain prior to its analysis with machine learning tools. We support that to achieve meaningful conclusions, the transformed data we analyze has to be as faithful as possible to the original recording, so that the transformations forced into the data due to restrictions in the T–F computation are not extended to the results of the machine learning analysis. Moreover, bioinspired computation such as brain–machine interface may be enriched from a more precise definition of neuronal coding where non-linearities of the neuronal dynamics are considered.

Medium spiny neurons activity reveals the discrete segregation of mouse dorsal striatum

“…To analyze the oscillatory activity, we used NA-MEMD algorithm ( Rehman and Mandic, 2010 ) together with Hilbert transform ( Huang et al, 1998 ). This algorithm is suited to decompose nonlinear nonstationary signals ( Alegre-Cortés et al, 2017 ; Alegre-Cortés et al, 2016 ; Hu and Liang, 2014 ; Mandic et al, 2013 ). Given the well-known nonlinear properties of neural oscillations ( Averbeck et al, 2006 ; Cole et al, 2017 ; Laurent, 1996 ; Shamir and Sompolinsky, 2004 ), the use of NA-MEMD leads to an increased detail of description when compared with traditional techniques ( Alegre-Cortés et al, 2017 ; Hu and Liang, 2014 ).…”

“…We used Noise-assisted Multivariate Empirical Mode Decomposition (NA-MEMD) algorithm ( ur Rehman and Mandic, 2011 ) together with Hilbert transform ( Huang et al, 1998 ) for the analysis of high-frequency oscillations both in LFPs and MSNS membrane potential, as well as the corticostriatal propagation of the SWO. Because neuronal oscillations are characterized by nonlinear properties, this algorithm is suited to decompose nonlinear nonstationary signals ( Alegre-Cortés et al, 2017 ; Alegre-Cortés et al, 2016 ; Hu and Liang, 2014 ; Mandic et al, 2013 ). The original EMD ( Huang et al, 1998 ) is a data-driven algorithm suitable for nonlinear and non-stationary signals that does not rely on any predetermined template.…”

Medium spiny neurons activity reveals the discrete segregation of mouse dorsal striatum

“…This algorithm is suited to decompose nonlinear nonstationary signals (Alegre-Cortés et al, 2018; Alegre-Cortés et al, 2016; Hu and Liang, 2014;Mandic et al, 2013). Given the well-known nonlinear properties of neural oscillations (Averbeck et al, 2006;Laurent, 1996;Shamir and Sompolinsky, 2004), the use of NA-MEMD leads to an increased detail of description when compared with traditional techniques (Alegre-Cortés et al, 2018;Hu and Liang, 2014). In our knowledge is the first time that it is used to analyse membrane oscillations of single neurons.…”

“…We used Noise-assisted Multivariate Empirical Mode Decomposition (NA-MEMD) algorithm (Ur Rehman and Mandic, 2011) together with Hilbert transform (Huang et al, 1998) for the analysis of oscillations both in LFPs and MSNS membrane potential. Because neuronal oscillations are characterized by nonlinear properties, this algorithm is suited to decompose nonlinear nonstationary signals (Alegre-Cortés et al, 2018;Alegre-Cortés et al, 2016;Hu and Liang, 2014;Mandic et al, 2013). The original EMD (Huang et al, 1998) is a data driven algorithm suitable for nonlinear and non-stationary signals that does not rely on any predetermined template.…”

Multiscale dynamics of interstimulus interval integration in visual cortex

“…We used an extension of the EMD algorithm[38] to study the T-F properties of the neural response. It has recently described [30,39] that the result of using EMD family of algorithms to study the oscillatory properties of spike trains improves the results obtained by means of using other traditional T-F techniques due to the presence of nonlinearities and nonstationarities in the signal [30–32,40].…”