Triple-kernel gated attention-based multiple instance learning with contrastive learning for medical image analysis.

Area of Science:

• Machine Learning
• Medical Image Analysis
• Deep Learning

Background:

• Multiple instance learning (MIL) is a supervised learning approach where data is organized in 'bags' of instances.
• Existing MIL methods face limitations in medical image analysis applications.
• Accurate and interpretable medical image classification remains a challenge.

Purpose of the Study:

• To propose a novel deep multiple instance learning model for enhanced medical image analysis.
• To overcome the limitations of current MIL approaches in the medical domain.
• To improve classification performance and provide explanatory information for disease models.

Main Methods:

• A novel triple-kernel gated attention-based MIL model incorporating contrastive learning.
• Representation extraction using a convolutional neural network with contrastive learning.
• Attention map generation using three distinct kernel functions to weigh instance importance.
• Attention-based MIL pooling for image aggregation and classification.

Main Results:

• The proposed model significantly outperforms state-of-the-art methods on binary and weakly supervised classification tasks.
• Demonstrated superior performance across diverse medical imaging datasets.
• Achieved more efficient classification results compared to existing approaches.

Conclusions:

• The novel deep MIL model offers a powerful tool for medical image analysis.
• The model provides enhanced classification accuracy and valuable explanatory information.
• This approach holds promise for improving diagnostic capabilities in various disease models.