Brain tumor segmentation and prediction on MRI images using deep learning network

Abstract

Brain Tumor is caused when the anomalousl cells that form within the brain and these could be of any size, shape in nature, so it is one of the difficult tasks to detect the presence of tumor. This could be found using MRI scans. In this paper, suitable algorithms have been developed to detect the MRI image has a brain tumor or not. The dataset used here has been taken from kaggle competition. Data augmentation is performed to maximize the data in dataset and this could results in having huge data. Since tumor area can overlap with non-tumor area of the MRI image, preprocessing steps is used to differentiate the images. So the proposed idea is to recognize tumors, this utilizes pre-processing strategies like filters, image enhancements, cropping, dilation, erosion, etc and for image classification pre-trained model InceptionResNetv2 is used as an CNN algorithm to detect whether the tumor is present or not. Various combination of pre-processing steps has been performed to find the effective pipeline for the classification. With the image pre processing techniques like cropped , median filter and CLAHE is gives a accuracy of 98.03% after the classification.