Comparative Analysis of Deep Learning Methods for Brain Tumor Detection

Abstract

The brain is one of the most important organs in the body. Nowadays, the number of people who lose their lives due to brain tumors is quite high. Additionally, there has been an increase in brain tumor cases worldwide due to climate and dietary habits, as well as the influence of environmental factors. Tumors that form in the brain have two types, benign and malignant. Although the risk of death is not very high for benign tumors, early detection is very important. These tumors can occur in any part of the brain and can be of different sizes. Since brain transplantation is not possible with today's technology, if the tumor is not treated early, there is a risk of losing the patient. Medical imaging systems are currently being used by expert physicians to detect brain tumors and allow for direct observation and analysis in a computer environment. Even if certain tests are performed by an expert, some tumors may go unnoticed. To prevent this, with the development of technology, AI-supported medical imaging systems produce supportive and accurate results to assist physicians in their decision-making processes. In this study, a Convolutional Neural Network (CNN) classifier model was used to detect tumors that had formed in the brain. In addition to this model, performance was tested between the ResNET50, VGG16, and VGG19 classifier models. As a result of the analysis, the developed CNN Classifier model achieved an accuracy of 94.70%, the ResNET50 model achieved an accuracy of 71.95%, the VGG16 model achieved an accuracy of 97.24%, and the VGG19 model achieved an accuracy of 92.38%. After determining the effective parameter values at the end of the study, it is recommended that deep learning models can be used in decision support systems for the analysis of medical images.