On-line identification of sensory systems using pseudorandom binary noise perturbations

Area of Science:

• Neuroscience
• Systems Biology
• Signal Processing

Background:

• Sensory systems, like the semicircular canal, provide crucial information about the body's orientation and movement.
• Understanding the linear characteristics of these systems is essential for interpreting neural signals.
• Previous methods for system identification were limited in their application to biological sensory outputs.

Purpose of the Study:

• To develop and validate a novel on-line technique for identifying linear system characteristics from sensory systems with spike train or analog voltage outputs.
• To apply this technique to the semicircular canal and assess its accuracy in modeling afferent responses.
• To investigate the impact of system variability on the identification process.

Main Methods:

• Developed a cross-correlation technique using pseudorandom binary white noise input to estimate unit impulse responses (UIRs).
• Corrected UIR estimates for input transducer response errors.
• Applied the technique to guitarfish horizontal semicircular canal afferent responses and known linear analog circuits for validation.
• Fitted UIRs to truncated exponential series using nonlinear regression to create parametric models.
• Quantified model accuracy by comparing experimental responses to predicted responses using percent mean-square-error (%MSE).

Main Results:

• The developed technique successfully generated non-parametric and parametric models of first-order unit afferent responses from the semicircular canal.
• Validation with linear analog circuits demonstrated the technique's ability to account for system variability.
• Analysis of 62 semicircular canal afferents yielded a low average %MSE of 10.2 +/- 5.9 SD%, indicating high accuracy.
• The technique effectively identified the linear characteristics of the semicircular canal afferent responses.

Conclusions:

• The developed on-line identification technique provides accurate inferences about the physiology of the semicircular canal.
• Linear system characteristics derived from afferent responses can be reliably used to understand semicircular canal function.
• This method offers a valuable tool for studying sensory system dynamics and neural coding.