An Averaging Ensemble Method for Multiclass Brain Tumor Classification Using Transfer Learning

A brain tumor is a disease that is caused by the abnormal growth of cells inside the brain or central spinal canal. It originates from the brain or, may spread to the brain from the other parts of the human body. While the exact causes of brain tumors remain unknown, environmental exposure and mutations in genes are the primary factors that significantly influence their progression. It is very difficult to diagnose whether the tumor is cancerous or not by analyzing the MRI images with bare eyes. For this, we need to use computer-aided diagnoses (CAD) to make the treatment easier. We have proposed an ensemble model that consists of the prediction of two deep learning models- Xception & InceptionV3 with accuracy of 97.89% and 91.56% respectively. We have used the data augmentation technique here. The proposed model gives an accuracy of 98.82%. We have also used the DenseNet121 for comparison purposes but among them, the proposed model performs better and gives high accuracy for the dataset contains about 7023 human brain MRI images which are categorized into 4 classes: glioma, meningioma, no-tumor & pituitary. The model averaging ensemble method results demonstrate that utilizing it can successfully categorize the samples that are incorrectly classified by the individual models.