Segmentation of Magnetic Resonance Brain Images Using Edge and Region Cooperation Characterization of Stroke Lesions

Yazid Cherfa,(1) Assia Cherfa,(1) Yacine Kabir,(1) Sammyh Kassous,(1) Assia Jaillard,(4) Michel Dojat,(3) and Catherine Garbay(2)

(1) Laboratoire LARIM, Dpt d’électronique, Université de Blida, Algérie

(2) Laboratoire CLIPS, Université Joseph Fourier, France

(3) Unité mixte INSERM/UJF 594, CHU, France

(4) CHU Michalon de Grenoble, Département Neurologie, France

Abstract: This paper describes an MR brain image segmentation system based on an edge and region cooperation. After an appropriate preprocessing using an improvement of image quality and brain insulation, we apply the cooperative segmentation to the preprocessed image. This segmentation is evaluated thereafter on reference image. The procedure is applied to several MR images from different subjects having undergone a cerebral vascular attack (stroke), this in order to extract the lesion and characterise them. Furthermore a comparison of our results with these obtained from manual measurements carried out by neurologists is presented.

Keywords: Magnetic resonance image, anisotropic filtering, brain segmentation, stroke, characterisation.

Received May 8, 2006; accepted September 14, 2006