Estimating nonstationary inputs from a single spike train based on a neuron model with adaptation.

Area of Science:

• Computational Neuroscience
• Neural Signal Processing

Background:

• Neuronal spike trains encode information about unobserved input signals.
• Existing state-space methods using the leaky integrate-and-fire (LIF) model track input parameters but face computational challenges with large datasets.

Purpose of the Study:

• To develop a more realistic and computationally feasible method for estimating neuronal input parameters from spike trains.
• To improve the accuracy of input parameter estimation by incorporating an adaptive threshold into the LIF model.

Main Methods:

• Augmented the leaky integrate-and-fire (LIF) model with an adaptive moving threshold for enhanced realism.
• Developed a practical method to transform instantaneous firing characteristics (firing rate, non-Poisson irregularity) back to input parameters.
• Utilized a computationally feasible algorithm for estimating firing characteristics from spike data.

Main Results:

• The proposed adaptive LIF model provides a more realistic estimation of neuronal input.
• The developed practical method efficiently transforms firing characteristics into input parameters, overcoming computational limitations of direct state-space methods.
• Validation on synthetic data demonstrated the effectiveness of the proposed methods.

Conclusions:

• The enhanced LIF model and practical transformation method offer a robust approach for inferring neuronal input dynamics.
• This work advances the computational neuroscience toolkit for analyzing neural spike train data.
• The methods are suitable for analyzing large-scale neural recordings.