Abstract
The computer-assisted study for improved deciphering imageries has been long-lasting topics in the field of medical imaging. Normally, various imaging techniques like ultrasound images, Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) can be used for assessment of the tumor for prostate, lung, brain, breast, liver, etc. For this research, we have used brain MRI images for identification of tumor. The vast volume of data which is generated by MRI scan foils physical tumor classification versus non-tumor for specific period. Brain tumor is almost common and violent sickness, which leads to a very shorter life. Hence instinctive and reliable classification scheme is important to prevent this rate of deaths of human being. Deep learning is speedily evidencing to be the state-of-the-art foundation, achieving enhanced performances by extracting features through a convolutional neural network (CNN) and then classifying via fully connected networks. In this research work, automatic classification of brain tumor is anticipated with the help of CNN because they are companionable to do the task of image classification, including image recognition, localization, segmentation, registration and detection. As segmentation gulfs the image into different regions on some similarity bases because of which we can abstract different features from the images. This research has increased the accurateness of classifying brain tumors. Experimentation results show that the designed model can detect tumor up to the accuracy of 95.71% on an online dataset. We have implemented the CNN architectures for detection and classification of tumors by the use of various models such as basic CNN architecture, VGG-16 architecture.
