An optimized support vector machine and deep convolutional neural network model to enhance brain tumor segmentation on magnetic resonance brain images

Abstract

Meningioma brain tumors are crucial and life killing disease among newlinethe other types of brain tumors. Hence, there timely and accurate detection and newlinesegmentation of tumor regions are very important to save the human life. In this newlinethesis, three different approaches are proposed to detect and classify the newlinemeningioma brain images from the non-meningioma brain images. newlineIn approach-1, rough k means clustering algorithm and Multi Kernel newlineSupport Vector Machine (MKSVM) algorithm is proposed to detect the tumor newlineregions in meningioma brain images. The purpose of these methods are to newlineprovide a Magnetic Resonance (MR) image segmentation, to raise accuracy for newlineclassification of the tumor feature direction and also maximizes and classified a newlineMR image. The preprocessing method is applied in this work to improve the newlineaccuracy of image segmentation and to reduce the noise. There are three steps newlineare followed in this work to achieve a effective results such as (i) The input of newlinebrain MRI images are preprocessed. (ii) The preprocessed images are delivered newlineto the feature extraction process then the feature extraction process is performed newlineby Improved Gabor Wavelet Transform (IGWT) (iii) Finally, feature values are newlinetransferred in to the clustering process for segmentation process. newline