Ensemble Model for Brain Tumor Classification from MRI Data

Abstract

Human brains are sophisticated and control the central nervous system. Brain abnormalities can disrupt organ processing and coordination. There are several ways to diagnose brain illnesses. Brain tumors in brain tissues are caused by uncontrolled cell proliferation. Brain tumors are among the deadliest disorders. Medical imaging is important in treating and mending crucial disorders. Medical brain imaging provides anatomical and functional information for diagnosis. This study classifies brain tumors using an ensemble of two pre-trained CNN RestNet-50 and GoogleNetclassifiers. We form the pre-trained CNN features into feature vectors for each MR brain image in the training and testing datasets. BRATS 2021 dataset is utilized in this research and model accuracy is compared with other pretrained models on these feature vectors to assess the performance of the classification system. Proposed Model (RES-GNET) achieves 97.9% accuracy for BRATS dataset on Kfold value 20. The model classifies medical pictures based on natural scene categorization. The suggested model has the maximum accuracy of 97.9%, surpassing other state-of-the-art algorithms. Future studies must integrate more recent datasets and use finer individual-level data. We need to conduct further research to understand the impact of healthcare access, treatment modalities, genetic predispositions, and socioeconomic factors on mortality outcomes.

Keywords: Brain tumor, CNN, Deep learning, Machine learning.

View PDF