Sparse Bayesian pMRI Reconstruction With Complex Bernoulli-Laplace Mixture Priors

Siwar Chaabene (1) , Lotfi Chaari (1, 2, 3) , Abdelaziz Kallel (4)

1 MIRACL - Multimedia, InfoRmation systems and Advanced Computing Laboratory
2 IRIT-TCI - Traitement et Compréhension d’Images
3 Toulouse INP - Institut National Polytechnique (Toulouse)
4 CRNS - Centre de Recherche en Numérique de Sfax

This paper presents a sparse Bayesian regularization technique for image restoration in parallel magnetic resonance imaging (pMRI). This technique is based on a hierarchical Bayesian model that solves the inverse problem of pMRI reconstruction by promoting sparsity using a Bernoulli-Laplace mixture prior. A Markov Chain Monte Carlo (MCMC) sampling technique is used to numerically approximate the target posterior. Our model allows handling complex-valued data. Promising results obtained on synthetic data demonstrate the performance of the proposed sparse Bayesian restoration model to provide accurate estimation of the target images.

Mots clés: Sparse Bayesian model; MCMC; Parallel MRI restoration

Domaines: Traitement du signal et de l'image [eess.SP]

Siwar Chaabene, Lotfi Chaari, Abdelaziz Kallel. Sparse Bayesian pMRI Reconstruction With Complex Bernoulli-Laplace Mixture Priors. 4th IEEE Middle East Conference on Biomedical Engineering (MECBME 2018), Mar 2018, Tunis, Tunisia. pp.193-197. ⟨hal-02947757⟩