A method of nonlinear analysis in the frequency domain

Area of Science:

• Neuroscience
• Systems Biology
• Signal Processing

Background:

• Nonlinear biological systems exhibit complex behaviors not captured by linear models.
• Analyzing these nonlinearities is crucial for understanding system function, particularly in neural pathways.
• Existing methods may struggle to fully characterize high-order nonlinear interactions.

Purpose of the Study:

• To develop a method for analyzing nonlinear biological systems using multi-sinusoidal input signals.
• To characterize nonlinear properties through frequency kernels and identify high-order interactions.
• To demonstrate the technique's utility in a biological system, specifically a cat retinal ganglion cell.

Main Methods:

• Input signals composed of a sum of numerous sinusoids to probe system nonlinearities.
• Analysis of system responses at harmonic and intermodulation frequencies.
• Derivation of frequency kernels analogous to Wiener kernels.
• Algorithm for phase variation to isolate high-order interactions.
• Application to data from a cat Y-type retinal ganglion cell.

Main Results:

• Frequency kernels effectively represent nonlinear system responses.
• Guidelines and examples for selecting input frequency sets are provided.
• A practical algorithm for separating high-order interactions is presented.
• Even-order nonlinear components dominate the response of Y-type retinal ganglion cells at high spatial frequencies.
• Fourth-order nonlinear components are detectable even at low contrast.

Conclusions:

• The developed method provides a powerful tool for analyzing nonlinear biological systems.
• The findings suggest an essential nonlinearity in the Y-type retinal ganglion cell pathway, with a singularity at zero contrast.
• The technique facilitates the detailed characterization of complex biological signal processing.