Tracking receptive field modulation during natural stimulation
Nicholas A. Lesica and Garrett B. Stanley
Division of Engineering and Applied Sciences
Harvard University
Traditional approaches to characterizing the stimulus/response mapping
in sensory systems make a number of simplifying assumptions: 1) the
stimulus is stationary and uncorrelated, 2) the mapping does not change
over time, and 3) the response of the neuron depends only on the
stimulus and is independent from one interval to the next. However,
characterizing the stimulus/response mapping in a natural setting
demands a more realistic model of sensory encoding in which stimuli of
arbitrary complexity are adaptively filtered into a history dependent
neural response. To identify the stimulus/response mapping in this
context, a new analytic approach must be developed. This poster
introduces a point process extended recursive least-squares (ERLS)
approach to receptive field estimation with the ability to: 1) estimate
RFs from responses to complex natural stimuli, 2) track adaptation of
receptive field properties during a single trial, and 3) capture the
behavior of a neuron more accurately by including history dependence in
a point process response model. This powerful approach allows us to
track RF modulation in retinal ganglion cells in response to changes in
contrast and investigate the nonlinearity of LGN responses to natural
scenes. The ERLS technique lends tremendous flexibility to experimental
design, which is essential for the investigation of sensory function in
the natural environment.