No abstract

Within-subject analysis in fMRI essentially addresses two problems, i.e., the detection of activated brain regions in response to an experimental task and the estimation of the underlying dynamics, also known as the characterisation of Hemodynamic response function (HRF). So far, both issues have been treated sequentially while it is known that the HRF model has a dramatic impact on the localisation of activations and that the HRF shape may vary from one region to another. In this paper, we conciliate both issues in a region-based joint detection-estimation framework that we develop in the Bayesian formalism. Instead of considering function basis to account for spatial variability, spatially adaptive General Linear Models are built upon region-based non-parametric estimation of brain dynamics. Regions are first identified as functionally homogeneous parcels in the mask of the grey matter using a specific procedure [Thirion, B., Flandin, G., Pinel, P., Roche, A., Ciuciu, P., Poline, J.-B., August 2006. Dealing with the shortcomings of spatial normalization: Multi-subject parcellation of fMRI datasets. Hum. Brain Mapp. 27 (8), 678-693.]. Then, in each parcel, prior information is embedded to constrain this estimation. Detection is achieved by modelling activating, deactivating and nonactivating voxels through mixture models within each parcel. From the posterior distribution, we infer upon the model parameters using Markov Chain Monte Carlo (MCMC) techniques. Bayesian model comparison allows us to emphasize on artificial datasets first that inhomogeneous gamma-Gaussian mixture models outperform Gaussian mixtures in terms of sensitivity/specificity trade-off and second that it is worthwhile modelling serial correlation through an AR(1) noise process at low signal-to-noise (SNR) ratio. Our approach is then validated on an fMRI experiment that studies habituation to auditory sentence repetition. This phenomenon is clearly recovered as well as the hierarchical temporal organisation of the superior temporal sulcus, which is directly derived from the parcel-based HRF estimates. © 2008 Elsevier Inc. All rights reserved.Keywords: Bayesian modelling; fMRI; Gamma-Gaussian mixture model; Detection-estimation; Markov Chain Monte Carlo methods; Bayes factor; Model comparison IntroductionSince the first report of the BOLD effect in human (Ogawa et al., 1990), functional magnetic resonance imaging (fMRI) has represented a powerful tool to non-invasively study the relation between cognitive stimulus and the hemodynamic (BOLD) response. Within-subject analysis in fMRI is usually addressed using a hypothesis-driven approach that actually postulates a model for the HRF and enable voxelwise inference in the General Linear Model (GLM) framework. In this formulation, the modelling of the BOLD response i.e., the definition of the design matrix is crucial. In its simplest form, this matrix relies on a spatially invariant temporal model of the BOLD signal across the brain meaning that the expected response to each stimulus is modelled by ...

No abstract

No abstract

No abstract

Different approaches have been considered so far to cope with the temporal correlation of fMRI data for brain activity detection. However, it has been reported that modeling this serial correlation has little influence on the estimate of the hemodynamic response function (HRF). In this paper, we examine this issue when performing a joint detectionestimation of brain activity in a given homogeneous region of interest (ROI). Following [1], we adopt a space-varying AR(1) temporal noise model and assess its influence, on both the estimation of the HRF and the detection of brain activity, using synthetic and real fMRI data. We show that this model yields a significant gain in detection specificity (lower false positive rate).

No abstract

No abstract