Classification of Brain Tumor using some approaches of Machine Learning

Abstract

Detecting brain tumors in their early stages is crucial. Therefore, prompt detection enables doctors to make timely and informed decisions. This dissertation focuses on developing and evaluating artificial intelligence techniques to identify the tumor and type of tumor using medical imaging and one of these medical imaging techniques to observe it is MRI. This work aims to improve the accuracy and efficiency of brain tumor detection . We have used artificial intelligence to detect the tumor and classify what type it is whether glioma, meningioma, pituitary or healthy brain, and from the artificial intelligence method we have used three machine learning techniques logistic regression, SVM, and decision tree for deep learning we have used CNN and Pre-trained model ResNet50. The machine learning methods have been trained after extracting features from brain tumor MRI with the SURF method and created BoF using two clustering techniques: K-means clustering and GMM clustering. The SVM that has been trained by features which have created using GMM clustering had the best performance in machine learning. The deep learning CNN and pre-trained have been trained on brain tumor MRI, the evaluation showed that the pre-trained had the best performance. Overall, the pre-trained model seemed to give the best accuracy by 0.976, and machine learning SVM had the best performance when the features were generated with GMM with an accuracy of 0.866.