Segmentation as an effective method of isolating a brain tumor on MRI

Abstract

Digital image processing, which ensues in many sides of life, is one of the areas that requires rapid development and improvement of existing algorithms, both for accuracy and completeness, and for reasons of speed and cost-effectiveness of both technical and software solutions. Medical application itself is the area where both precision in processing is important, as insufficient information affects the treatment protocol, and the cost for availability and widespread use. In this research, an algorithm for segmentation of digital MRI images of the brain is proposed in order to isolate the segment that contains the tumor. This algorithm is based on the sequential execution of the following steps: threshold Otsu’s method of binarization of the image, selection of brain and tumor tissues by morphological operations, segmentation by marked watershed, removal of the skull line and selection of the segment containing the tumor by an erosion. The verification did not reveal false-positive segmentation results, and the percentage of images correctly segmented to detect the tumor was 96.2%. It should be noted the high speed of the segmentation process obtained by the authors.