Brain Tumor Classification using Adaboost with SVM-based Classifier (ASVM) on MR Imaging

Abstract

Among malignant disorders, the brain tumor is the most threatening. When contrasted to other diseases, the odds of dying from a brain tumor are higher. Early identification of brain tumor is crucial to reduce the risk of fatality. Brain tumor segmentation, which entails the extraction of tumor patches from data’s, is top most difficult problems in the medical field. Due to the resemblance between cancerous and non - cancerous tissues and the wide variety of tumor appearances, this task is usually performed manually by medical experts, which is not always transparent. As a result, automating medical image segmentation remains a significant difficulty that has piqued the interest of a number of scholars in recent years. At this time, one of the biggest active study areas in the field of medical image processing is the detection and categorization of tumors. The goal of this research is to use an adaboosted SVM-based component classifier to construct a model for brain tumor detection and classification, i.e., to determine whether the tumor is cancerous or non-cancerous. A U-Net architecture is used initially for image segmentation, and then an adaboosted SVM classifier is used for classification. The purpose of using UNET is to increase parameter distribution accuracy and uniformity in the layers. Because of the difficulties in training SVM and the disproportion between diversity and accuracy over basic SVM classifiers, the Adaboost with SVM-based classifier (ASVM) is typically deemed to violate the Boosting principle. The Adaboost classifier in the study trains SVM as the basic classifier, which gradually decreases as the weight value of the training sample varies. On test data, the average classification accuracy is 98.2 percent, much outperforming SVM classifiers without Adaboost.