Hybrid algorithms for brain tumor segmentation, classification and feature extraction

The brain tumor is a cancerous disease due to the growth of abnormal cells in the human brain. It causes the death of many precious lives because of inaccurate calculation and identification of brain tumor. The average annual mortality rate of brain tumors in the United States between 2010 to 2014 was 4.33%, and almost 10,190 men and 7,830 women died this year from a brain tumor and the average survival rate in 5 years brain tumor is 36%. Much research has been done in the biomedical image processing field using computing concepts to segment and classifies brain tumors accurately. However, the diverse image content, occlusion, noisy image, chaotic object, nonuniform image texture, and other factors badly affect the performance of image clustering and segmentation algorithms. Therefore it is required to model an automatic image segmentation and classification algorithm. This research aims to segment brain tumors from MRI images using threshold segmentation and watershed algorithm and then classify brain tumors on features extracted (MSER, FAST, Harlick, etc.) through different classifiers. The proposed methodology includes image acquisition, image pre-processing, image segmentation, and feature extraction. Different classifiers are used to classify brain tumors from the datasets used accurately. The results indicate that the proposed mechanism enhances the detection of brain tumor images than the existing techniques by achieving more than 90% accuracy.