Brain Tumor Segmentation from Brain Magnetic Resonance Images using Clustering Algorithm
V. Malathy1, S. M. Kamali2

1V. Malathy, Assistant Professor, Department of Electronics and Communication Engineering, S R Engineering College, Warangal, India.

2S. M. Kamali, Assistant Professor, Department of Electrical and Electronics Engineering, Mahendra College of Engineering, Salem, India.

Manuscript received on 10 June 2019 | Revised Manuscript received on 17 June 2019 | Manuscript Published on 19 June 2019 | PP: 625-629 | Volume-8 Issue-8S June 2019 | Retrieval Number: H11060688S19/19©BEIESP

Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (

Abstract: The main purpose of this article is to ascertain the presence of the tumor as well as quantify it. This is achieved by using clustering-based methods of segmentation. Both clustering methods for K-means and FCM are used to segment the tumor. FCM clustering is better than K-means clustering due to its fuzzy assigning of the pixels to each cluster. This helped to isolate the tumor portion more effectively as pixels, previously not enlisted as tumor region, are detected to be tumor in FCM clustering. The paper‘s aim is also to find out if the tumor has reached critical stage or not. This is done by taking the scale of the MRI and using the number of pixels isolated. The other important aim of the paper is to remove noise effectively to get the best possible segmented output. Though median and mean filters are highly effective noise removal filters they do not aide in the removal of multiplicative noise. These filters only remove the additive noise present in the image and are not suitable for removing the multiplicative noise present. Therefore, discrete wavelet transformation is used. This helps to get better detection of multiplicative noise from the image as it works in the frequency domain. This fact is ascertained by the comparisons made with detection of the tumor from the original image and filtered images. For this paper, noise was added for it to be more visual though unnecessary. This fact is driven by the results obtained. The results clearly indicate that the filtered image gives better results.

Keywords: Magnetic Resonance Imaging, K-means Clustering, Fuzzy C-means Clustering, Brain Tumor.
Scope of the Article: Fuzzy Logics