Improving Detection and Classification Of Brain Tumors Using DenseNet201

Abstract

Brain tumors are among the leading causes of death worldwide. A brain tumor may originate in the brain or develop elsewhere in the body and metastasize to the brain, leading to secondary brain tumors. Thus, brain tumors can take many different forms. In this study, brain tumors were detected and classified based on magnetic resonance imaging (MRI) involving three different types of brain tumors and non-tumors. VGG-16, VGG-19 (Visual Geometry Group), DenseNet201 (Densely Connected Convolutional Networks), Inceptionv3, ResNet-50 (Residual Network with 50 layers), and EfficientNet-B0 were among the convolutional neural network (CNN) models that were employed and analyzed in order to determine the optimal model for detecting and classifying brain cancers. The best model was DenseNet201, which achieved accuracy, precision, and recall of 99.31%, 99.31%, and 99.25%, respectively.