Improved Brain Tumor Diagnosis through Feature Selection and Hybrid Ensemble Models

Brain and central nervous system tumors comprise more than 120 different types, making them a complex and significant health concern around the world. In particular, brain tumors are among the leading causes of cancer-related mortality. Early and accurate diagnosis plays an important role in improving patient survival rates. Although traditional machine learning algorithms have been used for brain tumor classification, they often face limitations in handling the complexity and diversity of tumor types. To overcome these challenges, our study proposes an advanced approach utilizing feature selection and ensemble learning techniques for enhanced brain tumor detection. We employed the minimum redundancy-maximum relevance method for reducing noise and improving model performance. Following feature selection, an ensemble learning strategy was implemented using Artificial Neural Networks, Random Forest and Back Propagation Neural Networks classifiers, combined with a voting mechanism for final classification. This approach not only improves diagnostic accuracy but also contributes to better clinical decision-making and patient outcomes. This ensemble approach leverages the strengths of each model, achieving an impressive accuracy of 98.27%, thus significantly improving classification reliability.