Medical Image Classification Based on Semi-Supervised Generative Adversarial Network and Pseudo-Labelling.

Area of Science:

• Medical imaging
• Artificial intelligence
• Machine learning

Background:

• Deep learning models require large labeled datasets for medical image classification.
• Medical data sharing restrictions and high labeling costs hinder deep learning applications.
• Existing methods struggle with limited labeled medical data.

Purpose of the Study:

• To develop a novel semi-supervised learning method for medical image classification.
• To leverage unlabeled medical images to improve model performance.
• To address the challenges of data scarcity and labeling costs in medical AI.

Main Methods:

• Proposed a novel method combining semi-supervised adversarial learning and pseudo-labeling.
• Incorporated unlabeled images into the model training process.
• Validated the method on ChestX-ray14 and BreakHis datasets.

Main Results:

• Achieved 93.15% accuracy for chest X-ray classification using only 30% of labeled samples.
• Outperformed current methods in multi-class breast cancer histopathological image classification with 96.87% accuracy.
• Demonstrated highly effective performance comparable to state-of-the-art results.

Conclusions:

• The proposed semi-supervised method effectively utilizes unlabeled data for medical image classification.
• This approach significantly improves diagnostic accuracy despite limited labeled data.
• The method shows promise for advancing AI applications in medical image analysis.