Comparison of Various Classification Techniques on Brain Tumor Detection

Abstract

Brain tumor is world’s most severe disease, and death rate has gradually increases. In this paper, we detect the brain different types of tumor via segmentation using neural network and classification using curvelet transform with unsupervised learning method. For that consideration, the process of new cells is replaced by old cells because of old cells are died first, but sometimes the old cells do not died and new cells are not able to replace other side; the new cells are often form a huge tissue and this called a tumor. We proposed the dynamic neural network with feed forwarded dynamic neurons for automatic image segmentation. The method, involvement of multiple neural layer architectures that particularly focus on extracting image features from image-related regions, besides having a positive effect against over fitting segmentation of MRI images by given the fewer number of weights in the network. Next, curvelet-based unsupervised learning technique for image classification, which is based on multistage and multidirectional transform that capturing the edge point in brain images. The edge pointes in an image are the significant information caring points, which are used to demonstrate better visual configuration of the image. Our approach achieved better results in brain tumor segmentation challenges on dataset by comparing with other existing methods. Our approach achieved better results in brain tumor segmentation challenges on BRATS 2019 dataset by comparing with other existing methods.