Evaluating the Performance of Deep Neural Networks in Brain Tumor Diagnosis

In computer-aided diagnosis (CAD) for medical applications, there is a serious problem in the classification of brain tumors. To prevent further issues, early fault discovery is essential. The area of modern technology is most actively developing is MRI scanning. It is difficult and time-consuming to identify a tumor from a variety of images because the tumor's size in the brain vary significantly for various patients as well as the minute characteristics of the tumor. This paper proposes a brain tumor classification model based on convolutional neural networks. The model recognizes three distinct categories of brain tumors, including glioma, meningioma, pituitary, and no tumor, from image data, through training with the suggested CNN model using transfer learning techniques that is VGG16 and EfficientNetB0. The VGG model converges with 91% accuracy whereas EfficientNetB0 converges with 94.8% accuracy. The proposed work uses the Kaggle databases of Brain Tumor Classification (MRI) in the testing stage to assess the model's performance after the CNN network model training is complete.