Design and Implementation of Hybrid Approach for Brain Tumor Detection and Demarcation

Abstract

Brain tumor is one of the most severe and often deadly disorders. The research of the image of the brain tumor has now received more and more attention. MRI is currently very useful for brain imaging without radioisotope injection. MRI is based on images that include a significant quantity of information, which produce diverse images by modifying the parameters. Because of the complicated structure of brain tumors, broken borders and outside variables, including noise, the inference of tumor and edema from brain magnetic resonance imaging (MRI) data remains challengeful. A powerful hybrid clustering technique combined with morphological procedures for the segmentation of brain tumors is proposed in this study in order to reduce noise sensitivity and increase segments stability. In this paper: First of all, Wiener adaptive filtration is used to de-noise, and morphology is employed for removing non-brain tissue, so successfully lowering the sensitivity of the technique to noise. Secondly, the clustering of K-means+ is merged with the C-means algorithm on the segment pictures of the Gaussian kernel. This grouping enhances not just the stability of the method, but also the clustering parameters' sensitivity. Lastly, the tumor images retrieved are processed utilizing morphological procedures and median filtering for the appropriate representation of brain tumors. The suggested technique has also been compared with other existing segmentation algorithms.