Automated Hybrid Model for Detection and Classification of Brain Tumor

Detection of brain tumors remains a critical challenge in the medical field. However, early brain tumor detection improves treatment options and enhances the likelihood of survival. Nevertheless, due to an extensive set of features in magnetic resonance images (MRI), manually checking them is a time-consuming and difficult task. To address these issues, brain tumor segmentation and classification must be done automatically. This research presents a novel framework to detect brain tumors through MRI using U-Net architecture and SVM classifier. The proposed research is based on pre-processing MRI that enhance and filter the images to improve contrast and eliminate noise. Moreover, segmentation of the MRI to extract the area of interest is performed through the modified U-Net architecture. After segmentation, a support vector machine (SVM) algorithm is employed to classify the normal and tumor-affected images. The proposed method is trained, validated, and tested on publicly available benchmarked BRATS 2019 and BRATS 2020 datasets. According to experimental findings, the proposed hybrid method achieved an overall accuracy of 98.95%, specificity of 98.13%, and sensitivity of 97.46%. This approach can be used in the medical field to assist clinicians and doctors dealing with a brain tumor diagnosis and treatment.