Multi-template approaches for segmenting the hippocampus: the case of the SACHA software
Résumé
The hippocampus has been shown to play a crucial role in memory and learning. Its volumetry is a well-established biomarker of Alzheimer’s Disease (AD) and hippocampal sclerosis in temporal lobe epilepsy (TLE). Manual segmentation being time consuming and suffering from low reproducibility,
robust automatic segmentation from routine T1 images is of high interest for studying large datasets. We previously proposed such an approach (SACHA, Chupin et al, 2007, 2009), based on competitive region deformation constrained by both anatomical landmarks and a single probabilistic template of 16 young healthy subjects registered using SPM5. The atlas being introduced as a soft constraint, robust results have been obtained in large series of patients with various pathologies. In recent years, multitemplate approaches have proven to be a powerful mean to increase segmentation robustness (Barnes et al, 2008) (Aljabar et al, 2009) (Heckemann et al 2006), more specifically for subjects with very large atrophy or atypical shapes (such as malrotations (Bernasconi et al, 2005) (Kim et al, 2012)).We propose here to evaluate the introduction of multiple-template constraints in SACHA.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...