A Study on Brain Tumor Extraction using Various Segmentation Techniques

Brain is the vital internal organ and the disease in brain affect the whole physiological system. The tumor is one of the brain abnormality normally arises in humans due to various reasons. The unrecognized and untreated brain tumor will lead to death. Hence, recently, a considerable number of programmed and semi-automated tumor recognition systems are proposed and implemented by the researchers. The proposed work aims to evaluate the performance of various segmentation techniques, such as Active-Contour (AC), Level-Set (LS), Watershed (WS) and Region-Growing (RG) methods existing in literature. This work considered the BRATS2015 benchmark brain MRI dataset for the evaluation. This work considered 150 numbers of images for the assessment (10 patients x 5 slices per patient x 3 views = 150). The proposed work is implemented on the two-dimensional (2D) slices of the brain, such as axial, coronal and Sagittal. The performance of the proposed methodology is confirmed based on the average segmentation accuracy attained during a comparative assessment with the ground-truth image. This work is implemented on 2D brain MRI slice without the skull section and the experimental results confirms that, proposed technique helps to attain the better accuracy on the axial-view image compared to other images.