Cancer Diagnosis Application Using Transfer Learning

Abstract

A brain tumor is the growth of abnormal cells in the tissues of the brain. It affects a large number of people of different ages worldwide. Magnetic Resonance Imaging (MRI) is the most operative and widely used technique for brain tumor detection as it provides better contrast images of the brain. However, the complexity level of the problem, manual classification process, requirement of skilled medical practitioners and a huge amount of MRI scan data are major factors thwarting timely classification of tumor vs. non-tumor. However, early detection of brain tumor is possible by applying machine learning (ML). Recently, deep learning automatic classification schemes based on Neural Networks (NN) are highly effective for classification tasks. A tumor diagnosis application is presented with VGG-19 based deep learning model by applying transfer learning of knowledge. The 5-fold cross-validation of model provided 88% accuracy along with 0.881 F1-score. The application could, could be utilized as a successful tool aid for radiologists in clinical diagnostics process.

Keywords

• Neural Network
• medical imaging
• brain tumour classification
• magnetic resonance imaging
• machine learning