Morphological- watershed algorithm for MRI brain tuoror segmentation

Abstract

Tumor in the brain is a deadly and intractable disease. Therefore, brain tumor detection and segmentation have played a vitally important role for helping doctors in the process of diagnosis and treatment. Moreover the earlier tumor is detected, the easier treatment is. However, there are some problems for detecting the brain tumor in the early stage, such as: the tumor is not clear, the quality of the MRI images is not suitable, or noise appears in the images. Therefore, this paper proposed a method of segmenting MRI brain tumor image based on the combination of Sobel filter, morphological operator and watershed algorithm. In particular, the Sobel mask will be used for tumor edge detection, then dilation operator is applied to link all dashed tumor boundaries, before the watershed algorithm is implemented to detect and segment the tumor regions. Finally, depending on the watershed transformed image, the tumors are clarified in the result image. With this proposed technique, the tumor segmentation processing is implemented automatically, the location and the shape of brain tumors can be simply detected visually and the time consumption for diagnosis is reduced.