Detection and Classification of Brain Tumor on MR Imaging using Deep Neural Network based VGG-19 Architecture

Abstract

: The massive growth of abnormal cell development in the brain region is known as a tumor. It is treated as a high prior disease in the modern medical domain, and it is difficult to cure. This type of tumor can be controlled only if it is diagnosed at an earlier stage. For making the accurate analysis and diagnosis process, the MR imaging tool is used by the radiologist. The exact portion of the tumor can be addressed by an MR image from the brain region. A deep convolutional neural network-based (DCNN) on Visual Geometry Group (VGG-19) architecture is proposed to detect the malignant portion in the brain region from the brain magnetic resonance imaging (MRI) dataset. The publically available BraTS dataset is used in our experimental study. The proposed DCNN uses a layer-based automatic segmentation and classification technique, and the hierarchy of the system is followed by, preprocessing, segmentation, feature extraction, and classification. A softmax classifier is used alongside the classification process, in order to classify the brain MR images efficiently. All together, obtained training and testing accuracy outcome of the proposed system is 99.2%, and the training and testing loss outcomes are 0.158 and 0.138 respectively.