Dementia-Agents: A Multi-Modal Multi-Agent System for Dementia Staging and Phenotyping

The paper introduces Dementia-Agents, a multi-modal multi-agent system designed for staging and phenotyping dementia by integrating diverse clinical assessments in real-world settings. The framework employs a three-step workflow involving a data agent for processing clinical records, multiple expert agents for generating predictions, and a coordinator agent for aggregating results, demonstrating improved diagnostic performance over existing multi-modal large language models and prior systems. This approach is significant for practitioners as it enhances the interpretability and accuracy of dementia diagnosis, addressing the complexities of syndrome-level assessments beyond traditional Alzheimer's-focused models.