A Novel Ensemble Model for Brain Tumor Diagnosis

Abstract

The spread of brain tumors resulted in numerous deaths, and cancer patients are still being treated. Four novel models are introduced and compared in this study. The best one is the PIEnsemble model, which was created to correctly identify and classify MRI images for brain tumor classification. The PIEnsemble primarily combines three deep learning techniques, ResNet Model, GoogleNet Model, and Inception Model, integrated with two dimensionality reduction techniques, PCA and ICA, which are used for feature extraction with a combination of Linear, Batch1Norm1D, and ReLU layers. Simulations in a series on two benchmarks datasets were run to show the improved effectiveness of the PIEnsemble model. The experimental results highlighted the improvements to the PIEnsemble classifier structure, which has the highest classification accuracy ratio and superior performance to other methods, with a 98.25% accuracy for Dataset 1 and a 98.75% accuracy for Dataset 2.