The Effective Study of Transfer Learning with VGG-16 on Detection and Classification of Brain Tumor

Abstract

Nowadays transfer learning is a popular area of research and it is giving a very huge impact particularly when dealing detection and classification of various classes of a specific disease with the aid of biomedical images. In this context, the proposed framework is based on brain MRI images for the detection and classification of various classes of brain tumors such as Meningioma, Glioma, and Pituitary tumors. The dataset utilized is openly available and it is popularly known as “Brain tumor dataset-Figshare”. It consists of all three classes of MRI images. The proposed framework is considered based on the concept of transfer learning with pre-trained CNN architecture, VGG-16. The proposed framework able to attain an overall accuracy of around 92% for the detection of the tumor, and the accuracy attained for the detected tumors to classify into meningioma, glioma, and pituitary tumors is 98.66%, 98.53%, and 99.54% respectively.KeywordsTransfer learningVGG-16CNN architectureBrain tumor