Abstract

Image segmentation is a process of partitioning an image into different subregions based on edge detection, area based or clustering based methods. Segmentation of brain MRI images is a challenging task. This paper provides a thorough analysis of different segmentation techniques with morphological operators for brain tumor detection. After segmenting the image, morphological operators are used to eliminate and add some pixels from tumor boundaries and to improve the performance of segmentation algorithm. Manual segmentation is used to construct the gold standard for comparing the segmented image. Comparison is performed using performance parameters such as dice, Jaccard coefficient, selectivity, recall and precision. The experimental results show that precision can be improved up to 85% in clustering-based segmentation and full selectivity can be achieved by combining segmentation techniques with morphological operation of erosion. The other performance parameters have also improved by applying erosion than dilation.