Abstract:

Abstract Analysis of Brain Tumor plays predominant role in detecting the tumor cells of the brain. At its most advanced stages, a brain tumor can be extremely fatal. It can quickly travel to other sections of the brain and harm healthy brain cells due to its uncontrollable reproduction. As a result, early detection is critical in the treatment of patients with the goal of increasing their life expectancy. However, because tumors have complicated characteristics in terms of appearance and limits, detecting them is a difficult and demanding task. For the diagnosis of brain cancers, magnetic resonance imaging (MRI) is widely employed, which necessitates segmenting large volumes of 3D MRI images, which is difficult to do manually. A modified version of VGG16 CNN and a sequential model were proposed for the automatic segmentation and detection of a brain tumor utilizing MRI images in this system. The suggested CNN model is compared to AlexNet, ResNet-50, VGG-16, and GoogleNet, which are all popular functional CNN models. Using 1030 brain MRI scans, it is able to achieve an overall accuracy of 98 percent and a cross entropy of 0.097. Using the Adam optimization approach, all of the key hyper parameters of CNN models are automatically designated. The proposed CNN models can be used to help physicians and radiologists validate their initial brain tumour screening with high accuracy and efficiency.

Keywords: