Improved Brain Extraction Tool using Marker-Controlled Watershed Segmentation

Abstract

MRI head scan images contains some non-brain tissues and it is necessary to remove these nonbrain tissues for further processing. The Brain Extraction Tool (BET) is one of the most popular brain segmentation technique which removes the non-brain tissues in MRI head scan volumes. It is a simple and robust brain extraction method. It is a boundary based method and uses deformable surface to detect brain boundaries. Due to its over smoothing capabilities it always under-segment the brain images by including more non-brain tissues in its final segmentation. The aim of this proposed method is to overcome the problems encountered with the existing BET method by using the marker-controlled watershed segmentation. The proposed improved brain extraction tool (IMBET) was tested with 20 volumes of brain images and its performance was compared with the popular existing methods BET, BSE, WAT, HWA and GCUT. The computed similarity measures Dice (D) and Jaccard (J) show that the proposed method outperform the BET and produced better brain extraction result than the existing methods.