Automatic Detection of Brain Tumor Using Deep Learning Algorithms

Brain tumor is the result of an abnormal growth of cells, which reproduce themselves in an uncontrolled manner. This type of tumour is diagnosed through Magnetic Resonance Imaging (MRI), which plays a significant role in segmenting the tumor region into different ways for performing surgical and medical planning assessment but the manual detection may lead to errors and it is a time consuming process. To overcome the problem, experts use various algorithms for automatic detection of the tumor region, which are based on deep learning algorithms. They are designed to train and tune millions of images within a short period of time. Further, this paper proposes different types of classification methods with a number of iterations are based on CNN architectures such as VggNet, GoogleNet and ResNet 50. For 60 iterations VggNet reports 89.33% accuracy, GoogleNet 93.45% and ResNet 50 96.50%. Finally, it is proved that ResNet 50 achieves better results than VggNet and GoogleNet with comparatively less time and better accuracy.