Author(s): Yazid Cherfa, Assia Cherfas, and Assia Jaillard
Article publication date: 2009-09-01
Vol. 27 No. 3 (yearly), pp. 103-113.
171
Keywords
Magnetic resonance image, anisotropic filtering, brain segmentation, stroke, characterization.
Abstract
In this paper, we present a semi-automatic method for extracting stroke lesions from 2D
MRI brain images using edge and region cooperation. Cooperative segmentation is applied after an
appropriate image pre-processing using an improvement of the image quality and skull stripping.
This segmentation is evaluated thereafter by a comparison to a reference image. To extract and
characterize lesions, we apply our method to several MRI images from different subjects who
experienced cerebral vascular attacks (strokes). We compare our results with those obtained through
manual measurements conducted by neurologists. We apply the results to reconstruct and visualise
3D strokes in the brain.