Automated segmentation of brain tumors in MRI using force data clustering algorithm

Abstract

In this paper, we present a novel automated method for detecting tumor location in brain magnetic resonance (MR) images, and identifying the tumor boundary. This method employs an unsupervised learning algorithm called Force for coarse detection of the tumor region. Once tumor area is identified, further processing is done in the local neighborhood of the tumor to determine its boundary. The Force method, which is based on the rules of electrostatics, is used for finding spatial clusters of high intensity in the 2D space of MR image. Further analysis of the identified clusters is performed to select the cluster that contains the tumor. This method outperforms many existing methods due to its accuracy and speed. The performance of the proposed method has been verified by examining MR images of different patients. © 2009 Springer-Verlag.