Comparison between convolutional neural networks (CNNs) and support vector machine algorithm (SVM) in classifying brain MRI tumors.

Abstract

Brain tumors are among the diseases that endanger human health. This disease not only harms people physically, but also puts financial and emotional obstacles to family and professional life. The key to early diagnosis of brain tumors is to draw conclusions from doctors' observation of MRI images of the brain. Brain tumors are divided into high-grade tumors and low-grade tumors. However, because each doctor has different experience and approaches to treating tumors, doctors have no way to have a uniform classification standard for brain tumors. In this research, a large number of brain MRI images were obtained from the 2019 Brain Tumor Segmentation challenge dataset. We process the brain MRI image data through a computer, and then train the data through a machine learning algorithm, which can classify tumors. brain effectively. In the experiment, convolutional neural network and support vector machine were used as the model training algorithm. The accuracy of the support vector machine on the test set was 84.3%, while the performance of the convolutional neural network on the test set was 77.6%. From the two experimental results, it can be concluded that the performance of SVM is better than that of CNN in both brain tumor classifications. In addition, in this study we also tried to improve the model performance through different model parameters. This research can be applied to the system for diagnosing brain tumors as well as other tumors. We hope that this paper will provide researchers with useful experience in this field.