Segmentation of Brain Tumor Magnetic Resonance Imaging using Convolutional Neural Networks

In today’s world, one of the most aggressive and common diseases that in its greatest rise leads to short life expectancy is known to be brain tumors. In the current work, for diagnosing brain tumors, we have used MRI images which generate enormous data which at a particular time impedes the tumor vs non-tumor’s manual classification. Along with some disadvantages of this technique, it has been observed that for a restricted number of images, it provides precise quantitative measurements. Thus, to prevent the human death rate, there is a need for an automatic and trusted classification scheme. Thus, we are proposing CNN classification for the automatic detection of brain tumors in this work. For exploring a small 3×3 kernel, a CNN-based automatic segmentation approach is suggested in this paper. Small kernels are used as they help in deeper architecture design, along with a positive influence against overfitting, as the network consists of a small number of weights.