Fiber Orientation Distribution from Non-Negative Sparse Recovery - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Fiber Orientation Distribution from Non-Negative Sparse Recovery

Résumé

The Fiber Orientation Distribution (FOD) [3] is a high angular resolution diffusion imaging (HARDI) model for robustly estimating crossing white-matter fiber bundles from q-ball acquisitions. However, its angular resolution depends on the spherical harmonic (SH) / tensor basis order, which implies a large number of acquisitions: 45, 66, 91 for typically used orders such as 8, 10, 12. Further, it is still necessary to compute the fiber orientations from the FOD. In the literature two ways have been adopted for this purpose: maxima detection and tensor decomposition. To overcome this two step approach (FOD estimation + fiber detection), we have proposed a novel FOD model and estimation method based on non-negative sparse recovery [1, 2]. The method has the following advantages: (i) it naturally estimates non-negative FODs, (ii) it computes both the FOD (tensor) and the fiber-orientations together – making tensor decomposition (which is NP-hard) or maxima detection unnecessary, (iii) it doesn’t require the number of fiber-compartments to be predefined and (iv) it can estimate very high order FOD tensors from a minimal number of acquisitions (20 or 30). We adopt this method for single shell data of this challenge.
Fichier principal
Vignette du fichier
Poster-Megherbi.pdf (330.48 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01104163 , version 1 (19-01-2015)

Identifiants

  • HAL Id : hal-01104163 , version 1

Citer

Thinhinane Megherbi, Aurobrata Ghosh, Fatima Oulebsir Boumghar, Rachid Deriche. Fiber Orientation Distribution from Non-Negative Sparse Recovery. MICCAI 2014 Workshop on Computational Diffusion MRI, Sep 2014, Boston, United States. ⟨hal-01104163⟩

Collections

INRIA INRIA2
224 Consultations
171 Téléchargements

Partager

Gmail Facebook X LinkedIn More