BNLoop-GAN: a multi-loop generative adversarial model on brain network learning to classify Alzheimer's disease.

Area of Science:

• Artificial Intelligence
• Neuroimaging
• Brain Informatics

Background:

• Current AI models for neuroimaging lack incremental learning capabilities, hindering continuous improvement.
• Alzheimer's Disease (AD) research benefits from AI but requires enhanced learning strategies.
• Batch training limitations in AI models impede effective analysis of complex brain data.

Purpose of the Study:

• To develop an AI model with incremental learning for analyzing multi-modal neuroimaging data.
• To enhance the classification of Alzheimer's Disease using continuous learning methodologies.
• To address limitations in existing AI models for brain network analysis.

Main Methods:

• Introduced the BNLoop-GAN (Loop-based Generative Adversarial Network for Brain Network) model.
• Utilized conditional generation, patch-based discrimination, and Wasserstein gradient penalty for learning brain network distributions.
• Developed a multiple-loop-learning algorithm for evidence combination and sample contribution ranking.

Main Results:

• Demonstrated effectiveness in classifying individuals with Alzheimer's Disease (AD) and healthy controls.
• Showcased improved classification performance using multi-modal brain networks and the BNLoop-GAN model.
• Validated the approach through various experimental designs and multi-modal data fusion.

Conclusions:

• The BNLoop-GAN model, combined with multi-modal brain networks and multiple-loop-learning, significantly enhances classification performance.
• Continuous learning strategies are crucial for advancing AI in neuroimaging and brain disease research.
• The proposed methodology offers a robust framework for evidence combination and fusion in brain informatics.