Segmentation is one of techniques used for classifying brain tissues in Magnetic Resonance Image (MRI) for identifying anatomical structures in the brain. The automated brain tumour segmentation remains challenging and computationally intensive because tumour appears in different size and intensity. In this paper, we have proposed a method for fast and automatic segmentation of tumour from Region of Interest (ROI) identified in MRI. ROI is a smaller portion of the image containing tumour. In the first step, tumour slices are identified using bilateral asymmetry property of the brain. In the second step, the ROI is identified using quadtree decomposition and similarity detection based on coefficient computed with gray level intensity histograms. In the third step, only the ROI is segmented using spectral clustering method rather than considering the whole image. Experimental results on real-world datasets are carried and compared with the recent existing works which show better results in terms of accuracy and less processing time for segmentation. © School of Engineering, Taylor’s University.