An Optimized Clustering Approach for Tumor Segmentation Using Local Difference of Intensity Level in MR Brain Images

Abstract

There are many computerized methods used to detect and identify brain tumor but tumor segmentation is still the most challenging task in medical image processing for designing an effective medical decision making system. In research and diagnostic studies performed on brain tumors, radiologists can use medical decision making system as a second reader in addition to his expert view on analyzing the brain images due to the complexity of brain structures. This study presents a new approach named LDI-Means algorithm (Local Difference in Intensity-Means algorithm) for image segmentation based on clustering technique by exploiting the difference in the intensity level of each pixel than another. The experimental results provided an approximate match with accuracy of 99.02% to the hand labeled images regardless the grade of glioma tumor, leading to faster and more precise method of brain tumor segmentation, detection and localization to ease patient management.