Accurate Detection of Brain Tumor Using Compound Filter and Deep Neural Network

The unexpected growth of nerves inside the human brain that interferes with the normal function of the brain is referred to as a brain tumor. Magnetic Resonance Imaging (MRI) is used to provide images of better resolution of the brain. This paper proposes a system that applies a compound filter, along with a Convolution Neural Network (CNN) and Support Vector Machine (SVM) for the detection of brain tumors in MRI. For finding tumors this proposed system has been divided into the following sections: preprocessing, segmentation, feature extraction and tumor detection. This system employs a compound filter for preprocessing that is made up of Gaussian, mean and median filters. Threshold and histogram-based techniques have been applied for image segmentation and Grey Level Co-occurrence Matrix (GLCM) for feature extraction. For tumor detection, the SVM and CNN classifiers were employed. CNN is a Deep Neural Network (DNN) based classifier. The tumor detection accuracy of CNN and SVM classifiers have been estimated at 98.06% and 93.28%, respectively. The proposed system concludes that the accuracy of CNN is superior to SVM.