Performance Analysis of Brain Tumor Detection Using Modified Adaboost Classification (MAC) Algorithm

Abstract

The abrupt growth of the cells in brain region forms the tumors. The brain tumors are the threatening diseases in persons all over the world. Detection of tumor regions in brain images is important for preventing death around the world. This paper uses Modified Adaboost Classification (MAC) algorithm for segmenting the tumor regions in brain images. The median filtering is applied on the brain image which detects and smoothes the noise contents in the image. This median filtering method splits the entire brain image into number of non-overlapping regions and then noise contents in each non-overlapping region are removed from the brain image in order to improve the smoothing process. This paper develops Modified Adaboost Classification (MAC) algorithm for the classifications of brain images for disease detection and diagnosis process. Further, the morphological mathematical equations are used to segment the tumor regions from the classified tumor brain image. The performance of the proposed MAC algorithm is analyzed with respect to accuracy and False Alarm (FA) rate. The brain image samples used in this paper are obtained from ‘Brainweb’ open access dataset. In this paper, total number of tumor affected brain image samples used is 189 and this proposed MAC algorithm correctly classified 186 samples and 3 samples are wrongly detected by the proposed algorithm. Hence, the accuracy of the proposed MAC algorithm for brain tumor detection is about 98.4% and the FA is about 1.6%. The proposed MAC algorithm obtains 98.4% of accuracy and 1.6% of FA on T1-WI images, obtains 97.8% of accuracy and 2.2% of FA on T2-WI images and obtains 98.4% of accuracy and 1.6% of FA on FLAIR images.