A Comparison Study of Different Classification Algorithm on Brain Tumor Segmentation

Abstract

A brain tumor is a tissue mass caused by aberrant cell proliferation in the brain. It is a collection of tissues that causes hormonal alterations and eventually death. In order to save human lives, brain tumors prognosis and prevention is a difficult task. The use of modern medical image processing approaches has made the identification of brain tumors more flexible in recent years. Due to the absence of ionizing radiations, the well-known approach for brain imaging is Magnetic Resonance Imaging (MRI). Existing brain image processing approaches have serious flaws, such as high false positive rates and low accuracy. We aimed to find the best brain tumors detection system by evaluating the performance of well-known classifiers in this research. To do this, we devise a perfect brain tumors detection system by overcoming the flaws that present in standard approaches. In the suggested method, the preprocessed region is segmented using morphological techniques and median filtering. Our suggested system employs a Genetic Algorithm (GA)-based feature optimization to achieve improved accuracy by picking the optimal subset from the input photos. Three well-known classifiers, Decision Tree J48, k-Nearest Neighbor (KNN), and Multi-Layer Perceptron, were used to infer these selected features (MLP). The performance of all of these classifiers was examined in the paper, and the specificity, sensitivity, accuracy, and error rates were used to determine the accuracy of the suggested brain tumors detection system.