Automated Brain Tumor Detection Using Machine Learning Algorithms on Magnetic Resonance Imaging Scans

The tumor is an excessive development of dangerous cells in any body part. Brain tumors can be of mainly two types: benign and malignant. The benign tumor has structural consistency and doesn’t contain active cells, while the malignant tumor has no structural consistency and contains active cells. Detection of tumors is often done from magnetic resonance images, which are obtained from scans performed by radiologists and medical experts. Accurate assessment of tumors from these images can aid in timely diagnosis and consequent treatment of patients. The primary objective of this study is to segment, detect, and classify images to determine whether they contain tumors. In this study, a dataset consisting of nearly 4600 images is used for classification using machine learning. There are five features, namely image, class, format, mode, and shape, while the target variable is a binary class, i.e., tumor and no tumor. The image classification was performed using various methods such as efficient logistic regression, decision tree, support vector machine, and artificial neural network, for which the obtained accuracies were 70.7%, 99.8%, 96.3%, and 96.8%, respectively. From the findings, it is observed that decision trees and artificial neural network techniques outperform other algorithms in detecting brain tumors. It is concluded that timely interventions using magnetic resonance image scans and machine learning techniques can aid in assessing the physiopathology and biochemical alterations of tumors and thereby aid in reducing the mortality risk due to tumors.