Brain tumor detection in MRI images using CNN and SVM

Abstract

During the past years, the healthcare sector has witnessed a new way of improvement due to Machine Learning. Among many deep learning systems implemented in the medical field, Convolution Neural Networks have the most significant impact in the classification domain. The objective of this thesis is to create a system consisting of a Convolutional Neural Network for semantic segmentation of brain tumors and enhance its results using Support Vector Machine. Brain cancer accounts for approximately 2% of all cancers and determining the genetics that underpin certain tumors, along with the localization of them, can aid in the fight against this deadly illness; moreover, human examination and expert diagnosis can be time consuming and not accurate since the human decision can be incorrect and the surgeon or expert cannot, alone, validate the tumor type but requires the opinions of other experts in the field. The goal of this thesis is to create a system consisting of a Convolutional Neural Network for semantic segmentation of brain tumors to detect the exact location of the brain tumor and then enhance its performance by adopting Support Vector Machine to remove any misclassification of the location of the brain tumor in MR images. The dataset used for training is from Kaggle year 2019 which was preprocessed and then augmented for better results.