Open Collections

How well does a linear model predict the responses of primary visual cortex neurons to a natural scene stimulus?

University of British Columbia

The goal was to test how well a linear model of the responses of neurons in area 18 of cat visual cortex, derived from recordings made in anaesthetized adult cats, predicts responses to natural scene stimuli. Methods: Estimates of the spatio-temporal receptive field profile of the neurons were obtained by reverse correlation to an m-sequence stimulus (Reid et al., 1997). The receptive field estimate, together with a non-linear response function, was then used to give the expected probability, or rate, of spike firing (Chichilnisky, 2001; Ringach & Malone, 2007) during a time-varying natural scene stimulus. The ability of the model to describe the responses was assessed by computing the correlation coefficient between the rates predicted by the model and those observed during stimulation with a natural scene (Willmore & Smyth, David & Gallant, 2005). For each LN functional model identified for all real A18 neurons using m-sequence responses, a Poisson spike generator was added (Heeger, 2000) to simulate ‘LNP’ responses to m-sequence and natural scene stimuli, and was used to assess the statistical significance of the results. Results: The LN model, with parameters derived from responses to m-sequence stimuli, was able to predict responses to m-sequence stimuli with fairly high reliability (correlation coefficients in the range 0.84 – 0.96). However the model was only able to weakly predict responses to natural scene stimuli. This result was confirmed by comparing the correlation coefficients between predicted and observed firing rates obtained for actual and for simulated responses to the natural scene stimulus; values ranged from 0.14 to 0.59, in marked contrast to the simulated ones ranging from 0.47 to 0.88. Reasons for the inability of the LNP model to predict responses to natural scene stimuli are discussed.

Text

2011-01-04

Attribution 3.0 Unported

http://hdl.handle.net/2429/30518

Neuroscience

University of British Columbia

UBCV

http://creativecommons.org/licenses/by/3.0/