Errors in estimation of the input signal for integrate-and-fire neuronal models

Area of Science:

• Computational Neuroscience
• Mathematical Biology
• Systems Neuroscience

Background:

• Stochastic neuronal models, like the integrate-and-fire (IF) type, are crucial for simulating neural dynamics.
• Accurate estimation of model parameters is essential for reliable predictions and experimental validation.
• The presence of a firing threshold in these models can introduce complexities in parameter estimation.

Purpose of the Study:

• To investigate the impact of the firing threshold on the estimation of input parameters for stochastic neuronal models.
• To quantify the systematic bias introduced by the firing threshold in different neuronal models.
• To provide guidance for experimental studies utilizing these models.

Main Methods:

• Analytical derivation of bias formulas for the randomized random walk and perfect integrator models.
• Monte Carlo simulations to analyze bias in the leaky integrate-and-fire (LIF) model.
• Comparison of the derived bias with other estimation errors.

Main Results:

• A systematic bias is identified in parameter estimation due to the firing threshold.
• Analytical expressions for this bias are derived for specific neuronal models.
• Monte Carlo simulations confirm the presence and magnitude of bias in the LIF model.
• The bias is shown to be a significant factor compared to other estimation errors.

Conclusions:

• The firing threshold introduces a systematic error in parameter estimation for stochastic neuronal models.
• This bias must be accounted for in experimental neuroscience to ensure accurate model interpretation.
• The findings are applicable to various integrate-and-fire neuronal models, including the LIF model.