Deep Transfer Learning Approaches in Performance Analysis of Brain Tumor Classification Using MRI Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Pathology

Background:

• Brain tumor classification is crucial for effective patient treatment and recovery.
• Magnetic Resonance Imaging (MRI) is a preferred non-ionizing radiation imaging modality for brain pathology.
• Deep learning, particularly Convolutional Neural Networks (CNNs), has significantly advanced automated medical image analysis.

Purpose of the Study:

• To perform a comparative analysis of transfer learning-based CNN models for brain tumor classification.
• To evaluate the performance of VGG-16, ResNet-50, and Inception-v3 for automatic brain tumor prediction from MRI.
• To identify the most effective pretrained CNN model for brain tumor localization and classification.

Main Methods:

• Utilized a dataset of 233 MRI brain tumor images.
• Applied transfer learning with pretrained CNN models: VGG-16, ResNet-50, and Inception-v3.
• Focused on the VGG-16 model for brain tumor detection and classification.

Main Results:

• The VGG-16 pretrained CNN model showed highly adequate results for brain tumor prediction.
• Demonstrated an increase in the accuracy rate for both training and validation phases using VGG-16.
• Comparative performance analysis indicated VGG-16's effectiveness in this task.

Conclusions:

• Transfer learning with pretrained CNNs, specifically VGG-16, is effective for automated brain tumor classification from MRI.
• The VGG-16 model offers a promising approach for improving diagnostic accuracy in neuro-oncology.
• Further research can leverage these findings for enhanced clinical decision support systems.