Watershed Transform Based on Clustering Techniques to Extract Brain Tumors in MRI

Abstract

In this work, watershed transform method was implemented to detect and extract tumors and abnormalities in MRI brain skull stripped images. An adaptive technique has been proposed to improve the performance of this method.Watershed transform algorithm based on clustering techniques: K-Means and FCM were implemented to reduce the oversegmentation problem. The K-Means and FCM clustered images were utilized as input images to the watershed algorithm as well as of the original image. The relative surface area of the extracted tumor region was calculated for each application. The results showed that watershed trnsform algorithm succeedeed to detect and extract the brain tumor regions very well according to the consult of a specialist doctor after viewing the resultant images. The adaptive technique, watershed based on clustered segmented image, improved the performance of the watershed transform and reduced the oversegmentation problem, and the utilizing of bilateral smoothing improves this result.