A Deep Learning-Based Multi-Model Approach for Brain Tumor Detection in MRI Images

Abstract

A brain tumor is a serious type of tumor that can be life-threatening. The most common tumor sites include the brain and nervous system tissues. Early and accurate diagnosis is crucial for brain tumor treatment. This study evaluated different deep learning models for the automatic classification of brain tumors using magnetic resonance imaging (MRI) data. Magnetic resonance imaging (MRI) provides very effective data for brain tumor diagnosis. This data significantly contributes to physicians' treatment process and helps achieve more successful outcomes compared to traditional treatment methods. Four different brain tumor types (glioma, pituitary, meningioma, and no tumor) were targeted in the classification process. The classification data were compared during the training and testing phases for a total of six different models commonly used in deep learning-based approaches: VGG16, ResNet-50, InceptionV3, Xception, MobileNetV2, and EfficientNet-B0 The model performance was evaluated using metrics such as accuracy, accuracy rate, loss function, and confusion matrix. Experiments revealed that the MobileNetV2 model achieved a classification accuracy of 90% compared to other models. This result demonstrates that lightweight models, particularly suitable for use in mobile and low-resource systems, can also be effective in complex medical imaging problems.