Spectral Clustering of Time-Intensity Signals: Application to Dynamic Contrast-Enhanced MR Images Segmentation
Résumé
In this paper, we propose to apply a spectral clustering approach to analyse dynamic contrast-enhance magnetic resonance (DCE-MR) time-intensity signals. This graph theory-based method allows for grouping of signals after space transformation. Grouped signals are then used to segment their related voxels. The number of clusters is automatically selected by maximizing the normalized modularity criterion. We have performed experiments with simulated data generated via pharmacokinetic modelling in order to demonstrate the feasibility and applicability of this kind of unsupervised approach.
Domaines
Ingénierie biomédicale| Origine | Fichiers produits par l'(les) auteur(s) |
|---|
