BRAIN TUMOUR DETECTION USING SUPER PIXEL BASED SPECTRAL CLUSTERING
Publication Date : 31/05/2019
A brain tumour or intracranial neoplasm occurs due to abnormal cell growth within the brain. It varies in different shape, size and intensity. Here, superpixel based spectral clustering method is been used for the segmentation of brain tumour from MRI. MRI (Magnetic Resonance Imaging) is used to create high quality clusters. The main problem during spectral clustering is that it suffers from dense similarity matrix construction. In the proposed work, the drawback of dense similarity matrix construction can be overcome in two steps (i.e., identification of region of interest (ROI) and then doing spectral clustering from obtained ROI). FCM and GMM algorithms are used to perform brain tumour segmentation. The dice score values of edema and tumour core regions are calculated from the MRI. From the results, it can be observed that the proposed method gives better result than the existing.
No. of Downloads :