Identification automatique des régions d'intérêts sur des images tomographiques rénales

Djamal Boukerroui (1) , Wala Touhami (2) , Jean-Pierre Cocquerez (1)

1 Heudiasyc - Heuristique et Diagnostic des Systèmes Complexes [Compiègne]
2 ENIT - Ecole Nationale d'Ingénieurs de Tunis

We propose, in this paper, an original approach in a statistical framework, for fully automatic delineation of kidneys (healthy and pathological) in 2D CT images. Our approach has two main steps: a localisation step followed by a delineation step. The localisation step is guided by a statistically learned prior spatial model in one hand and a grey level prior model in a second hand. The second step, utilizes the localisation results in order to precisely delineate the kidney's regions using a set of learned IF-THEN rules. The proposed approach is tested on clinically acquired images and promising results are obtained.

Domaines: Traitement des images [eess.IV]

Djamal Boukerroui, Wala Touhami, Jean-Pierre Cocquerez. Identification automatique des régions d'intérêts sur des images tomographiques rénales. Traitement du Signal, 2009, 26 (3), pp.239-254. ⟨hal-00429764⟩