Multi-Classification of Brain Tumors via Feature Level Ensemble of Convolutional Neural Networks

Abstract

The brain tumor is considered to be one of the deadliest diseases that can lead to cancer. Early and appropriate diagnosis can prevent its heinous consequences which are not often possible by solely depending on the manual detection process. A substantial result can be achieved using Convolutional Neural Network (CNN) in automating the process of detecting and classifying different types of of brain tumors. In this paper, a new ensemble architecture that consists of 3 individual efficient CNN models is proposed. The proposed feature level ensemble of 3 CNN models increases the model robustness and efficacy to a great extent. Further, Analysis of Principal Component (PCA) is done for feature or dimensionality reduction which also improves the performance of the model considering execution time and accuracy. The Magnetic Resonance Imaging (MRI) dataset that is used contains total 3064 brain images. It includes images of brain tumors and they are categorized as- meningioma, glioma, and pituitary. Our proposed model shows a prominent performance that outperforms other existing models along with the pre-trained models obtaining an average validation accuracy of 98.37%.