A Comparative Study to Detect Tumor in Brain MRI Images using Clustering Algorithms

The identification of the tumor area in the magnetic resonance images(MRI) by radiologists or experts is a tedious and time-consuming task. This task requires high accuracy, and that comes with experience and knowledge. With the growth in the information technology, medical imaging field is also reducing the complexities and increasing the accuracy in diagnosis. To increase the efficiency and accuracy of the Fuzzy K-Mean clustering, K-Mean clustering, and Birch clustering algorithms, these algorithms are incorporated with contour-based cropping, pre-processing, and post-processing. In pre-processing, anisotropic diffusion filter, non-local means filter, and Gaussian filter are used and compared. In post-processing, erosion and median filter are used. The experimental results show the significance by comparing the quality parameters with the state-of-the-art methods. Index Terms—MRI image, Tumor, Clustering algorithm, morphological, and contour detection