New paper: MAGDA: Multi-agent guideline-driven diagnostic assistance

[maykcaldas]

Paper: MAGDA: Multi-agent guideline-driven diagnostic assistance

Authors: David Bani-Harouni, Nassir Navab, Matthias Keicher

Abstract: In emergency departments, rural hospitals, or clinics in less developedregions, clinicians often lack fast image analysis by trained radiologists,which can have a detrimental effect on patients' healthcare. Large LanguageModels (LLMs) have the potential to alleviate some pressure from theseclinicians by providing insights that can help them in their decision-making.While these LLMs achieve high test results on medical exams showcasing theirgreat theoretical medical knowledge, they tend not to follow medicalguidelines. In this work, we introduce a new approach for zero-shotguideline-driven decision support. We model a system of multiple LLM agentsaugmented with a contrastive vision-language model that collaborate to reach apatient diagnosis. After providing the agents with simple diagnosticguidelines, they will synthesize prompts and screen the image for findingsfollowing these guidelines. Finally, they provide understandablechain-of-thought reasoning for their diagnosis, which is then self-refined toconsider inter-dependencies between diseases. As our method is zero-shot, it isadaptable to settings with rare diseases, where training data is limited, butexpert-crafted disease descriptions are available. We evaluate our method ontwo chest X-ray datasets, CheXpert and ChestX-ray 14 Longtail, showcasingperformance improvement over existing zero-shot methods and generalizability torare diseases.

Link: https://arxiv.org/abs/2409.06351

Reasoning: produce the answer}. We start by examining the title and abstract for any mention of language models. The title "MAGDA: Multi-agent guideline-driven diagnostic assistance" does not explicitly mention language models, but it does suggest a system involving multiple agents for diagnostic assistance.

Next, we look at the abstract. The abstract mentions "Large Language Models (LLMs)" and describes a system that uses multiple LLM agents augmented with a contrastive vision-language model to assist in medical diagnosis. The focus is on using these language models to follow medical guidelines and provide diagnostic support.

Given that the abstract explicitly discusses the use of large language models and their role in the proposed system, it is clear that the paper is about a language model.