Comparative analysis of machine learning techniques on the BraTS dataset for brain tumor classification

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare

Background:

• Accurate brain tumor classification from MRI scans is crucial for patient outcomes.
• Machine learning (ML) and deep learning (DL) show potential, but their comparative effectiveness is unclear.

Purpose of the Study:

• To evaluate and compare the performance of various ML and DL models for brain tumor classification.
• To determine the optimal approach for automated brain tumor diagnosis using MRI data.

Main Methods:

• Utilized the BraTS 2024 dataset for model evaluation.
• Assessed traditional ML (Random Forest) and DL models (CNN, VGG16, VGG19, ResNet50, Inception-ResNetV2, EfficientNet).
• Applied preprocessing techniques to enhance model performance.

Main Results:

• Random Forest achieved the highest accuracy at 87%.
• Deep learning models ranged in accuracy from 47% to 70%.
• Traditional ML surpassed advanced DL methods in this specific classification task.

Conclusions:

• Model selection and parameter tuning are critical for automated brain tumor diagnosis.
• Random Forest demonstrated superior performance over DL models in this study.
• Optimized ML approaches can enhance tumor classification accuracy and efficiency for diagnostic systems.