Reshaping Neuroimaging by Utilizing Transfer Learning and VGG Architecture and Ensemble with DANN Model

This research paper puts major emphasis on the importance of early detection of anomalies in the human brain and performing classification of three types of brain abnormalities such as Pituitary Tumor, Glioma Tumor and Meningioma Tumor. The dataset consists of these three types of brain tumor affected MRI images of brain with normal brain MRI images as well. To achieve this we have opted for the VGG16 architecture as a base model achieving an accuracy of 92.8% and to further enhance it, integrated Domain Adversarial Neural Networks (DANN) architecture, achieving 73% of accuracy on the entire dataset. The ensemble of both of these models has contributed to the performance and various domain adaptability of the model.