Isolation of Relevant Visual Features from Random Stimuli for Cortical Complex Cells

Abstract

A crucial step in understanding the function of a neural circuit in visual processing is to know what stimulus features are represented in the spiking activity of the neurons. For neurons with complex, nonlinear response properties, characterization of feature representation requires measurement of their responses to a large ensemble of visual stimuli and an analysis technique that allows identification of relevant features in the stimuli. In the present study, we recorded the responses of complex cells in the primary visual cortex of the cat to spatiotemporal random-bar stimuli and applied spike-triggered correlation analysis of the stimulus ensemble. For each complex cell, we were able to isolate a small number of relevant features from a large number of null features in the random-bar stimuli. Using these features as visual stimuli, we found that each relevant feature excited the neuron effectively in isolation and contributed to the response additively when combined with other features. In contrast, the null features evoked little or no response in isolation and divisively suppressed the responses to relevant features. Thus, for each cortical complex cell, visual inputs can be decomposed into two distinct types of features (relevant and null), and additive and divisive interactions between these features may constitute the basic operations in visual cortical processing.

Spike-triggered characterization of excitatory and suppressive stimulus dimensions in monkey V1

“…But if the neural response depends on more than a single axis within the stimulus space, the STA will provide an insu cient and possibly misleading description. A number of authors have suggested the natural extension of examining higher-order statistical properties (and in particular, the covariance) of the spike-triggered ensemble of stimuli [2,3,8,9,11]. The idea is simple and intuitive: if the neural response is deter-mined by a projection onto a low-dimensional subspace (within the space of all stimuli), an analysis of the spike-triggered covariance might allow us to recover this subspace.…”

Blue-Yellow Signals Are Enhanced by Spatiotemporal Luminance Contrast in Macaque V1

“…Recently developed data-analysis tools have provided new ways to characterize neurons that combine inputs nonlinearly (de Ruyter van Steveninck and Bialek 1988;Paninski 2003;Rust et al 2004; Schwartz et al 2001;Sharpee et al 2004;Simoncelli et al 2004;Touryan et al 2002). Here we use one of these techniques to reveal a surprising nonlinear computation performed by blue-yellow neurons in V1.…”

“…While the linearity assumption is justified for many V1 neurons, it is clearly inappropriate for others (Conway 2001;Hanazawa et al 2000;Hubel and Wiesel 1968;Lennie et al 1990;Vautin and Dow 1985). For example, a V1 neuron that responds to S cone stimulation, but only when L-and M-cone excitations are appropriately balanced, does not integrate cone inputs linearly and thus cannot be characterized with coneisolating stimuli (Hanazawa et al 2000).Recently developed data-analysis tools have provided new ways to characterize neurons that combine inputs nonlinearly (de Ruyter van Steveninck and Bialek 1988;Paninski 2003;Rust et al 2004; Schwartz et al 2001;Sharpee et al 2004;Simoncelli et al 2004;Touryan et al 2002). Here we use one of these techniques to reveal a surprising nonlinear computation performed by blue-yellow neurons in V1.…”

Analysis and interpretation of quadratic models of receptive fields

“…(B) In physiological experiments (i) In physiological experiments, random quadratic forms can be generated by bootstrapping. The spikes can be shuffled 6,10 or the entire spike train can be shifted relative to the stimulus sequence 9,12 (this second possibility is to be preferred if the input stimuli have a temporal structure, i.e., a non-zero autocorrelation) and the same RF estimation procedure used to generate the units under consideration can be applied to the resulting data. The new randomly generated quadratic forms are compatible with the distribution of the input data and with the total number of spikes elicited in the neuron under consideration.…”

“…Quadratic forms are used in experimental studies as quadratic approximations to the input-output function of neurons and can be derived from neural data as Volterra/Wiener approximations up to the second order [3][4][5][6][7][8][9][10][11][12][13] . In addition, several theoretical studies have defined quadratic models of neuronal RFs either explicitly 2,14,15 or implicitly as neural networks [16][17][18] .…”