Segmentation of Ischemic Stroke Area from CT Brain Images

Due to its wide availability, speed, and price, Computer Tomography (CT) without contrast injection has been the first imaging modality for the evaluation of acute ischemic stroke. However, the early signs of brain ischemia are subtle and not visible on CT scan. For this regard, we propose, in this paper, an approach for ischemic detection based on contrast enhancement of CT brain images. Our method has three stages: preprocessing, image enhancement and classification. First, skull bones and noise are removed; then a multi-scale contrast enhancement algorithm based on Laplacian Pyramid (LP) is developed to enhance the contrast of the brain CT images for identification of hypo-attenuation of acute stroke. Finally, "fuzzy c means" classification is applied to extract the ischemic area from normal tissues. The execution of our method gives impressive results.