A Hybrid Model for Brain Tumor Classification via Deep Feature Extraction and Optimized Classifier Weighting

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate brain tumor classification is vital for timely treatment and diagnosis.
• Automated methods are needed to improve efficiency and accuracy in medical image analysis.

Purpose of the Study:

• To develop a novel multiclass brain tumor classification approach.
• To enhance classification accuracy by combining deep features with ensemble learning.

Main Methods:

• Utilized deep features extracted from the pre-trained VGG-16 model for Magnetic Resonance Images.
• Employed Support Vector Machine (SVM) and K-Nearest Neighbours (KNN) classifiers.
• Implemented majority voting to combine SVM and KNN decisions for improved prediction.

Main Results:

• The proposed ensemble model achieved 99.54% accuracy on the figshare brain tumor dataset.
• Demonstrated superior performance compared to individual SVM and KNN classifiers.
• Showcased improved robustness and minimized misclassification errors.

Conclusions:

• The combination of deep feature extraction and ensemble learning offers a reliable tool for multiclass brain tumor diagnosis.
• The proposed method surpasses the state-of-the-art in brain tumor classification.
• Highlights the potential of AI in advancing medical image analysis for oncology.