Mixed Effect Modelling of Single Trial Variability in Ultra-High Field fMRI

TL;DR

This paper introduces a mixed effect statistical model for ultra-high field fMRI data that estimates haemodynamic responses and trial variability, improving analysis of brain activity responses to stimuli.

Contribution

It extends existing models by estimating both response functions and variability without some previous assumptions, using maximum likelihood.

Findings

Response varies significantly with task and time.

Model effectively captures single-trial variability.

Application to motor cortex activity demonstrates practical utility.

Abstract

Neuronal brain activity in response to repeated stimuli can be perceived using functional magnetic resonance imaging (fMRI). In this paper, we develop a statistical model for fMRI data that estimates both the associated haemodynamic response function and the within and between trial variability through maximum likelihood estimation. We discuss our results in the context of other model-driven approaches, extending models already popular in the literature, while removing the need for some of their assumptions. We consider an application to the motor cortex activity caused by a subject pressing a button and observe that the response changes significantly with task and through time.