Efficient Feature Extraction and Selection Strategies for Brain Tumor Diagnosis with MRI Imaging

Abstract

This paper introduces a novel approach for feature extraction and classification of brain tumor regions in MRI images. Feature extraction is a critical step in medical imaging, as it converts raw data into a more manageable form, facilitating effective decision-making and pattern classification. The proposed method combines intensity, texture, and shape-based features to classify brain regions into categories such as white matter, gray matter, and both normal and abnormal areas. By integrating multiple feature types, the model enhances the ability to distinguish between different brain structures and tumor regions. The model achieves an impressive accuracy of up to 95%, significantly outperforming traditional classification models, such as KNN (69.88%), Random Forest (66%), and SVM (77.59%). This demonstrates the reliability and effectiveness of the proposed method in accurately detecting and classifying brain tumors, showing its potential for real-world clinical applications. The results highlight the model’s robustness and promise for improving diagnostic accuracy in medical imaging