Support vector machine classifier optimized with seagull optimization algorithm for brain tumor classification

Abstract

SummaryMagnetic resonance imaging (MRI) is a practical tool for diagnosing tumors in human brain. In this work, MRI images are analyzed to find tumor‐containing regions and classify these regions into three different types of tumor: meningioma, glioma, and pituitary. Several methods were utilized previously for brain tumor detection, but no one method classify the brain tumor accurately and also it takes high computation time. To overwhelm these issues, a support vector machine optimized with seagull optimization algorithm (SOA) is proposed for brain tumor classification (SVM‐SOA‐BTC). The brain MRI images are gathered via Brats image dataset. Then the images are preprocessed using Savitzky–Golay denoising method. Then residual exemplars local binary pattern based feature extraction is utilized for extracting radiomic features. Then, the extracted features are fed to SVM classifier for classifying the brain tumors, like normal and abnormal. Then the weight parameters of the SVM are optimized using the SOA. The simulation is implemented on MATLAB. Then the proposed SVM‐SOA‐BTC method achieves 33.78%, 19.69%, and 11.62% higher accuracy; 30.62%, 25.05%, and 9.10% higher F‐score compared with existing methods, like Deep‐CNN‐DSCA‐BTC, CNN‐WHHO‐BTC, and AFDNN‐FLA‐BTC respectively.