Classification Of Brain Tumor Using Resnet50

Abstract

Brain tumor is a deadly cancer type; early and precise identification is significant
for the medication process. In the past decades, the brain tumor is detected using computer
aided systems i.e. MRI Scan. The Analysis of brain tumor from MRI images has become an
efflorescent research area in the domain of medical imaging systems. The MRI images have an
advantage of high image resolution and it is also independent of ionizing radiation. Image
processing concepts can visualize the different anatomy structure of the tumor by locating the
tumorous part and examining its intensity. MRI images play a vital role in radiology making
the process more enduring and tedious with the increase in data. These MRI images must be
processed with the utmost care since a small glitch in the outcome of MRI images is fatal. The
functioning of the nervous system is inversely proportional to the growth rate of the brain
tumor. To overcome this problem, Convolutional Neural Network approach for brain tumor
classification is implemented here. The Deep neural network is able to pull out required
features from the input and tumors are classified. Initially, the MRI images are pre- processed,
RGB converted, image augmentation is performed and fed into the neural network. By using
ResNet architecture an accuracy of 98% is achieved.