Fitting Intracranial multimodal ERPs to an exploratory, hierarchically arranged neural mass model.
Poster Presentation

Rosalyn Moran
School of Electronic, Electrical and Mechanical Engineering, University College Dublin, Ireland

Richard Reilly
School of Electronic, Electrical and Mechanical Engineering, University College Dublin, Ireland

Sophie Molholm
Nathan Kline Psychiatric Research Institute

John Foxe
Nathan Kline Psychiatric Research Institute

     Abstract ID Number: 83
     Full text: Not available
     Last modified: March 16, 2006
     Presentation date: 06/18/2006 4:00 PM in Hamilton Building, Foyer
     (View Schedule)

Abstract
A recent neural mass model of ERP generation (David, Friston et al. Modelling event-related responses in the brain. NeuroImage 25, 2005.) has identified different ERP patterns dependent on the type of hierarchy in which a given cortical area is embedded. These hierarchies include (i) bottom-up processing with intrinsic excitation, where the feedforward structure may extend for one to many areas, (ii) top-down processing from high-level cortical areas and (iii) lateral processes with excitation from anatomically local neuronal pools.

Here, in the area of the lateral superior parietal lobule Event Related Potentials have been extracted for unimodal audio and visual, and multimodal AV behavioural, button-press response tasks using intracranial EEG. This study presents a methodology where this neural mass model may be fitted to these ERPs, for the purpose of examining multimodal integration.

Results show that pertinent ERP features, including important late components, are best fitted to real data by the optimisation of delay and area connectivity parameters. Model parameters that describe intrinsic area connections and operation (e.g. max EPSP, max IPSP, maximum firing rate) are less significant when attempting to fit the model set to real intracranial EEG.

This work is a necessary precursor to developing an optimisation scheme for multimodal, data-driven models.