Automatic clustering based approach for brain tumor extraction

Abstract

The brain cancer is a deadly disease affecting almost 1.7% of world’s population with top mortality rate. The basic cause of brain cancer is the abnormal growth of brain tissues in the early stage. Further, these tissues are converted into tumors. Early detection and exact location of brain tumor can assist for further therapies of brain. In this paper a computer vision-based approach is developed for exacting an exact location of the brain tumors from the MRI images. The method starts from skull stripping for removal of the outer portion of brain and segregate the white matter. Further communication with local agent (CLA) clustering technique is applied followed by morphological post processing methods for extraction of tumor regions from white matter regions of the brain. The method is tested on a publicly available MRI dataset. Quantitative and qualitative measures show that the proposed method achieves an accuracy of 99.64% as compared to others.