Novel Approach Using Transfer Deep Learning for Brain Tumor Prediction

Abstract

A brain tumor refers to an abnormal collection or aggregation of cells in the brain that has the potential to be life-threatening owing to the cells’ capacity to penetrate and metastasize to organs that are nearby. It is possible to save lives by making a correct diagnosis of this potentially fatal condition. Within the last several years, there has been a noticeable increase in the functionality of deep learning applications. As a result, improving the module’s architecture leads to better approximations in the monitored configuration. Through the provision of trustworthy datasets, the categorization of tumors via the use of deep learning algorithms has successfully achieved significant progress. The purpose of this article is to use transfer module algorithms for the prediction of brain tumors. These modules include MobileNet, VGG19, InceptionResNetV2, Inception, and DenseNet201. The suggested module uses three main optimizers: Adam, SGD, and RMSprop. The simulation findings indicate that the pre-trained MobileNet module with the RMSprop optimizer outperformed all other evaluated modules. In addition to having the shortest amount of time required for computing, it obtained an accuracy of 99.6 %, a sensitivity of 99.4 %, and a specificity of 100 %.