Brain Tumor Classification by Ensemble Learning Utilizing Wavelet Transformation-based Statistical Features

Brain tumors are regarded as highly dangerous and severe conditions, frequently leading to a substantial reduction in life expectancy. Accurate identification of brain tumors is essential for developing successful therapies that can increase patients' chances of survival. The process of manually examining and interpreting extensive MRI data can be both complex and labor-intensive. As a result of this, there is a pressing need for a trustworthy machine-learning method for accurately identifying brain tumors, and during the last decade, many approaches have been implemented. This research presents a novel method to identify as abnormal any given MR brain image. The proposed method begins with the application of DWT to extract features from images. To find the most pertinent features, PCA is then applied as a feature selection method. These selected features are classified using both a SVM with a polynomial kernel and an Ensemble of Decision Trees (EoDT), with k-fold cross-validation employed to improve the model's generalizability. Our proposed methods demonstrate enhanced performance, with the EoDT model attaining the maximum classification accuracy of 99.95%. The SVM model obtains an accuracy of 90.61%, while the combined model reaches an accuracy of 94.20%. The experimental data indicate that the method is both effective and efficient, making it suitable for brain tumor classification. It can assist doctors in diagnosing whether a patient is normal or abnormal to varying degrees.