BRAIN TUMOUR SEGMENTATION USING K-MEANS CLUSTERING

Mohammed Hameedullah Sharief, S.Yallamandaiah

Abstract: Medical image processing plays an important role for finding the various types of diseases in real world. Tumour is one of the disease like a cancer can leads to death and it can be detected with the help of image processing. Magnetic resonance imaging MRI image is taken as input image it is resized to 500X500 i.e. five hundread rows and five hundread coloumns and skull stripping is done by morphological operations by converting image to binary based upon ostus thresholding and this image is cleaned using erosion and gaps are filled if any through filling operation. This binary image is overlaid on original image and this image is skull erased image. Then feature extraction is done by calculating the entropy feature by sliding small size window and then feature vector is constructed and given as input to the k-means clustering, it cluster tumour in one region and non tumour in another region. K-means is an example of exclusive clustering in which data belongs to only one cluster. Here two clusters are taken and centroids are obtained from two clusters. Distance is calculated from each pixel value to centroids. After this validation is done with the help of confusion matrix by giving ground truth through region polygon tool which can be used for creating mask from original image then this image is converted to vector along segmented image and both are compared in confusion matrix. This confusion matrix will create a table of different parameters through which accuracy, error rate, specificity and precision are calculated

Keywords: Feature Extraction, K-Means Clustering, Region Polygon, Confusion Matrix.

DOI: https://doi.org/10.15623/ijret.2016.0509004