Brain Tumour Segmentation And Classification Using U-Net And Fully Convolutional Neural Network

  • ^Mr. Ashok V, Arjun M, Ajith Kumar P.P, Andrew L

Abstract

Brain tumour is an abnormal growth of living cells within the brain. This mass of tumour grows within the skull, due to which normal brain activity is affected. If brain tumour is not detected in earlier stages, it can take away the person’s life. Detection of brain tumour is a challenging task due to the complex structure of the brain which may vary from person to person. MR (Magnetic resonance) image can be a very useful tool to spot brain tumour. But accurate tumour area segmentation can be a complicated and time consuming task. Our application uses U-net based CNN (Convolutional Neural Network) running on PyTorch (AI and Computer Vision Library for Python) for accurate tumour segmentation from a given brain MR image. The detected boundaries of the tumour are used to determine the area, intensity, shape and the location of the tumour, which are then fed into another Fully Convolutional Neural Network for classification of the tumour into one of its major types – meningioma, glioma and pituitary tumour.

Published
2020-06-01