Comparing integrate-and-fire-like models estimated using intracellular and extracellular data.

Liam Paninski, Jonathan Pillow, and Eero Simoncelli

Neurocomputing 65: 379-385.

Presented at
Computational Neuroscience 2004, Baltimore, MD

We have recently developed a maximum-likelihood (ML) method for estimating integrate-and-fire-based stimulus encoding models that can be used even when only extracellular spike train data is available. Here we derive the ML estimator given the full intracellular voltage trace and apply both the extracellular-only and intracellular method to responses recorded in vitro, allowing a direct comparison of the model fits within a unified statistical framework. Both models are able to capture the behavior of these cells under dynamic stimulus conditions to a high degree of temporal precision, although we observe significant differences in the stochastic behavior of the two models.
Reprint (500K, pdf)  |  Related work on estimation of neural models  |  on integrate-and-fire models |  Liam Paninski's research