Early detection and classification of brain tumors is very important in clinical practice. In recent decades, deep learning has gained more interest in various fields like image classification, self-driven cars, natural language processing, and healthcare applications. It solves the complex problems in more effective and efficient manner. It is a subset of layers which comprises of convolutional neural network, activation function, and decision-making layers. In this article, AlexNet, GoogleNet, and ResNet101 networks are used to classify the MRI images into four classes, e.g., normal, glioma, meningioma, and pituitary tumors, which have been carried out.

Dataset consists of 120 MRI images with four different classes. The pre-trained networks are used to classify the class in three different ways. In the first case, 10 sample images are considered and its prediction rate and training and validation time are recorded. Similar in the second and third methods, 20 and 30 images are used to evaluate the metrics. The result concluded that, for more images, processing time is increased, yielding better accuracy. The highest accuracy of 91 is achieved by the ResNet101 network with the processing time of 6 minutes. Researchers can still reduce this processing time and increase the accuracy rate by incorporating the image augmentation techniques to the raw MRI images.