Abstract :This paper proposes an automatic method to detect stroke region using Computed Tomography (CT) images. identifying the edge in image processing is one of the essential techniques in segmentation ,separation and understanding the position of special parts of the picture . Presence of noise and structural dispersion caused by local weakness of medical images is one of the most important reasons that today's operations are not able to identify the edge of images correctly. Ant colony optimization as one of the recent heuristic approaches is applied to the 98 brain CT scans from stroke subjects and tested by Support vector machines (SVM) classification .The results show a good improvement in accuracy of stroke diagnosis.