Advancing brain tumor classification: A robust framework using EfficientNetV2 transfer learning and statistical...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Brain tumor diagnosis is critical for patient outcomes but faces challenges in accuracy and efficiency with current methods.
• Accurate classification of brain tumors is complex due to location and the need for precise treatment planning.
• Existing classification techniques often lack the required accuracy or efficiency for timely diagnosis.

Purpose of the Study:

• To develop a novel, highly accurate, and efficient approach for brain tumor classification using deep learning.
• To leverage transfer learning with the EfficientNetV2b0 architecture for enhanced feature extraction from medical images.
• To improve upon traditional methods in terms of classification accuracy, efficiency, and training speed.

Main Methods:

• Utilized a Convolutional Neural Network (CNN) based on the EfficientNetV2b0 architecture.
• Employed transfer learning, pre-trained on extensive datasets, to extract relevant image features.
• Implemented efficient preprocessing and data augmentation techniques for medical image analysis.

Main Results:

• Achieved a remarkable classification accuracy of 99.16% for brain tumors.
• Demonstrated high precision, recall, and F1 scores, indicating robust diagnostic performance.
• Comparative analysis showed superior efficacy over state-of-the-art CNN architectures like InceptionResNetV2 and Deep CNN.

Conclusions:

• The proposed EfficientNetV2-based model offers a robust and highly accurate solution for automated brain tumor classification using MRI scans.
• The approach significantly enhances diagnostic accuracy and efficiency, showing potential for clinical application.
• Findings contribute to advancing brain tumor diagnosis and improving patient outcomes through AI-driven insights.