Abstract:
  Modelling the haemodynamic response in functional magnetic resonance   (fMRI) experiments is an important aspect of the analysis of  functional neuroimages. This has been done in the past using   parametric response function from a limited family. In this   contribution, we adopt a semi-parametric approach based on finite   impulse response (FIR) filters. In order to cope with the increase in  the number of degrees of freedom, we introduce a Gaussian process  prior on the filter parameters. We show how to carry on the analysis  by incorporating prior knowledge on the filters, optimising  hyper-parameters using the evidence framework, or sampling using a   Markov Chain Monte Carlo (MCMC) approach.  We present a comparison of our model with standard haemodynamic response  kernels on simulated data, and perform a full analysis of data   acquired during an experiment involving visual stimulation.

  Keywords: Neuroimaging, Haemodynamic response, fMRI, FIR filters,
  Smoothness prior, Tikhonov regularisation, Evidence, Markov Chain, Monte Carlo.