Brain Tumor Classification using Deep Learning

A brain tumour is an abnormal aggregation of cells within the human brain. The skull is incredibly robust and encloses the brain. Any expansion inside such an area might cause issues. Brain tumors can be of two types i.e. cancerous (malignant) or non cancerous (benign). Brain cancer is a fatal disease and it affects all the diagnosed people severely. Accurate classification of brain tumor helps in early diagnosis of brain cancer, which increases the survival rate of brain cancer patients. It is difficult to evaluate the magnetic resonance imaging images manually. Therefore there is a requirement of fast and accurate computerized methods for tumor diagnosis. In this work we are using a deep learning model based on convolutional neural networks to identify and classify brain tumor and its type from MRI images of patients using publicly available dataset in kaggle. In this paper, we are using a VGG16, Convnet and Efficiennet- B0 models for deep learning multi image classification which classifies input image into one of four types like meningioma, glioma, pituitary and no tumor. The dataset includes 5712 and 1311, training MRI and testing MRI images respectively. The models VGG16, Convnet and Efficientnet-B0 achieves 96.27%, 84.13%, 92.5% accuracy respectively, a significant performance with a best overall accuracy of 96.27% provided by VGG16 model.