An improved convolutional neural network-based model for detecting brain tumors from augmented MRI images

Abstract

Identifying and categorizing a brain tumor is a crucial stage in enhancing knowledge of its underlying mechanisms. Brain tumor detection is one of the most complex challenges in modern medicine. There are a variety of diagnostic imaging techniques that may be used to locate malignancies in the brain. MRI technique has the unparallel image quality and hence serves the purpose. Deep learning methods put at the forefront have facilitated the new paradigm of automated medical image identification approaches. Therefore, reliable and automated categorization techniques are necessary for decreasing the mortality rate in humans caused by this significant chronic condition. To solve a binary problem involving MRI scans that either show or don’t show brain tumors, we offer an automatic classification method in this paper that uses a computationally efficient CNN. The goal is to determine whether the image shows brain tumors. We use the Br35H benchmark dataset for experimentation, freely available on the Internet. We augment the dataset before training to enhance accuracy and reduce time consumption. The experimental evaluation of statistical measures like accuracy, recall, precision, F1 score, and loss suggests that the proposed model outperforms other state-of-the-art methods.