Automated Brain Tumor Detection Using Hybrid Machine Learning Models

Abstract

Brain tumors are among the most dangerous and life-threatening neurological diseases, and early detection is essential for successful treatment and improved patient survival. Magnetic resonance imaging (MRI) is the most widely used modality for the rapid diagnosis of brain tumors, but the accurate segmentation and interpretation of MRI images remain challenging. Deep learning (DL) has recently brought significant progress in the identification and categorization of brain tumors. This paper proposes a hybrid machine learning model for brain tumor detection in which a pretrained MobileNetV2 convolutional neural network (CNN) extracts deep features from MRI images and a linear support vector machine (SVM) classifies them into four classes: glioma, meningioma, pituitary tumor, and no tumor. The performance of the model is reported with class-wise precision and recall and a confusion matrix.