Brain Tumor Detection Using Artificial Convolutional Neural Networks

A tumor is a mass of abnormal cells that accumulate forming a tissue. These abnormal cells feed on the normal body cells and destroy them and keep growing bigger. One of these tumors is Brain tumor. It affects the nervous cells, brain cells, glands and membranes that surrounds the brain. A tumor can be diagnosed via imaging and pathology. Brain tumor is imaged with MRI (Magnetic Resonance Imaging), giving a cross-section images of the brain. In this paper, we have constructed a model based on Artificial Convolutional Neural Networks that takes these magnetic resonance images and analyze them with mathematical formulas and matrices operations. This neural network gives us the probability of how likely the existence of tumor in the brain, and had trained over magnetic resonance images, the diversity of images was 155 healthy brains and 98 with tumor. In total the dataset contains of 253 magnetic resonance images. It was expanded, using data augmentation, to become 14 times larger. The model gave us excellent results of predicting the existence of a tumor which reached 96.7% in validation data and up to 88.25% on test data.