Brain tumor classification using GAN-augmented data with autoencoders and Swin Transformers.

Area of Science:

• Medical Image Analysis
• Artificial Intelligence
• Oncology

Background:

• Brain tumor classification is a complex medical imaging task with significant diagnostic challenges.
• Existing deep learning models often struggle with data imbalance and limited feature extraction, impacting accuracy.

Purpose of the Study:

• To develop a novel deep learning model for accurate brain tumor classification.
• To address data imbalance and enhance feature extraction in medical image analysis.

Main Methods:

• A hybrid deep learning model combining Swin Transformer and AE-cGAN augmentation was proposed.
• AE-cGAN was used for synthetic data generation to improve dataset diversity and model generalization.
• Swin Transformer was employed to capture intricate local and global dependencies in medical images.

Main Results:

• The model achieved high accuracy rates of 99.54% and 98.9% on two public datasets.
• Significant improvements in classification, sensitivity, and specificity were observed compared to state-of-the-art methods.
• The approach effectively mitigated data imbalance and feature extraction limitations.

Conclusions:

• The proposed deep learning model demonstrates superior performance in brain tumor classification.
• The integration of Swin Transformer and AE-cGAN effectively addresses key challenges in medical image analysis.
• Future work includes clinical deployment and application to diverse medical imaging tasks.