Tumor Segmentation and Detection using Convolutional Neural Network

Tumor is an old age disease, higher life expectancy rate, and population boom are contributing to the higher number of patients. Rare and aggressive tumors can have a lot of swelling associated with them. Some tumors may grow at a slower rate and may not be visible and grow over a long period of time. It becomes difficult to perform manual segmentation on MR images in a given period of time and highly dependent on the expertise and experience of the operator. In this paper, authors have proposed to detect tumor using Convolutional Neural Network (CNN) and U-Net based deep convolutional networks. The implementation is on Pytorch an open source machine, learning library developed by the Facebook artificial intelligence research group. Pytorch is basically the combination of two languages, python, and torch. Authors have used three major libraries named torch.nn, torch.autograd and torch.vision which provided automatic differentiation for all the operations on tensor. It basically returns the 2D tensor representing the identity matrix. Experimentation performed is to achieve a higher degree of accuracy in tumor segmentation. The results have been compared with existing literature and found satisfactory up to a certain level.