Conference Paper

FMRI analysis through Bayesian variable selection with a spatial prior

Dept. of Stat., Univ. of Illinois at Urbana-Champaign, Urbana, IL, USA
DOI: 10.1109/ISBI.2009.5193147 Conference: Biomedical Imaging: From Nano to Macro, 2009. ISBI '09. IEEE International Symposium on
Source: IEEE Xplore


This paper presents a novel spatial Bayesian method for simultaneous activation detection and hemodynamic response function (HRF) estimation of functional magnetic resonance imaging (fMRI) data. A Bayesian variable selection approach is used to induce shrinkage and sparsity, with a spatial prior on latent variables representing activated hemodynamic response components. Then, the activation map is generated from the full spectrum of posterior inference constructed through a Markov chain Monte Carlo scheme, and HRFs at different voxels are estimated non-parametrically with information pooling from neighboring voxels. By integrating functional activation detection and HRFs estimation in a unified framework, our method is more robust to noise and less sensitive to model mis-specification.

Full-text preview

Available from:
  • Source
    • "Moreover, FIR-based inference may lack robustness and often yields disrupted shapes that are difficult to interpret, due to over-fitting problems. Hence, to cope with this issue, temporal [22] [12] [36] [6] but also spatial regularization [21] [54] [11] [34] [42] [17] [28] [14] [7] [53] have been introduced in models to improve estimation accuracy. Such estimation methods can be categorized according to three main criteria: i.) the HRF model they rely on (parametric [14] [30] or FIR [12] [11] [1] [6] [35] [53]), ii.) the type of inference that they perform (univariate [25] [14] [1] vs. multivariate [53] [51]) and iii.) the adaptability degree of regularization they introduce, namely the supervised [30] vs. unsupervised [14] [53] [51] setting of hyper-parameters. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Inter-subject fMRI analyses have specific issues regarding the reliability of the results concerning both the detection of brain activation patterns and the estimation of the underlying dynamics. Among these issues lies the variability of the hemodynamic response function (HRF), that is usually accounted for using functional basis sets in the general linear model context. Here, we use the Joint Detection-Estimation approach (JDE) [35,51] which combines regional nonparametric HRF inference with spatially adaptive regularization of activation clusters to avoid global smoothing of fMRI images. We show that the JDE-based inference brings a significant improvement in statistical sensitivity for detecting evoked activity in parietal regions. In contrast, the canonical HRF associated with spatially adaptive regularization is more sensitive in other regions, such as motor cortex. This different regional behavior is shown to reflect a larger discrepancy of HRF with the canonical model. By varying parallel imaging acceleration factor, SNR-specific region-based hemodynamic parameters (activation delay and duration) were extracted from the JDE inference. Complementary analyses highlighted their significant departure from the canonical parameters and the strongest between-subject variability that occurs in the parietal region, irrespective of the SNR value. Finally, statistical evidence that the fluctuation of the HRF shape is responsible for the significant change in activation detection performance is demonstrated using paired t-tests between hemodynamic parameters inferred by GLM and JDE.
    Full-text · Article · Jun 2013 · NeuroImage

  • No preview · Chapter · Jun 2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: One of the major objectives of fMRI (functional magnetic resonance imaging) studies is to determine subject-specific areas of increased blood oxygenation level dependent (BOLD) signal contrast in response to a stimulus or task, and hence to infer regional neuronal activity. We posit and investigate a Bayesian approach that incorporates spatial dependence in the image and allows for the task-related change in the BOLD signal to change dynamically over the scanning session. In this way, our model accounts for potential learning effects, in addition to other mechanisms of temporal drift in task-related signals. However, using the posterior for inference requires Markov chain Monte Carlo (MCMC) methods. We study the properties of the model and the MCMC algorithms through their performance on simulated and real data sets.
    Full-text · Article · Sep 2014
Show more