[论文解读] A Survey on Medical Large Language Models: Technology, Application, Trustworthiness, and Future Directions

Lei Liu, Xiaoyan Yang+7 more|arXiv (Cornell University)|Jun 6, 2024

Machine Learning in Healthcare被引用 20

一句话总结

This survey traces the evolution of Medical Large Language Models (Med-LLMs) from general LLMs to medical-specific systems, detailing technologies, applications, trust, and future directions. It covers data, algorithms, evaluation, and ethical/regulatory considerations for Med-LLMs in healthcare.

ABSTRACT

With the advent of Large Language Models (LLMs), medical artificial intelligence (AI) has experienced substantial technological progress and paradigm shifts, highlighting the potential of LLMs to streamline healthcare delivery and improve patient outcomes. Considering this rapid technical progress, in this survey, we trace the recent advances of Medical Large Language Models (Med-LLMs), including the background, key findings, and mainstream techniques, especially for the evolution from general-purpose models to medical-specialized applications. Firstly, we delve into the foundational technology of Med-LLMs, indicating how general models can be progressively adapted and refined for the complicated medical tasks. Secondly, the wide-ranging applications of Med-LLMs are investigated across various healthcare domains, as well as an up-to-date review of existing Med-LLMs. The transformative impact of these models on daily medical practice is evident through their ability to assist clinicians, educators, and patients. Recognizing the importance of responsible innovation, we discuss the challenges associated with ensuring fairness, accountability, privacy, and robustness. Ethical considerations, rigorous evaluation methodologies, and the establishment of regulatory frameworks are crucial for building trustworthiness in the real-world system. We emphasize the need for ongoing scrutiny and development to maintain high standards of safety and reliability. Finally, we anticipate possible future trajectories for Med-LLMs, identifying key avenues for prudent expansion. By consolidating these insights, our review aims to provide professionals and researchers with a thorough understanding of the strengths and limitations of Med-LLMs, fostering a balanced and ethical approach to their integration into the healthcare ecosystem.

研究动机与目标

• 解释从通用 LLMs 到医疗特定的 Med-LLMs 的历史发展，以及推动这一转变的技术。
• 总结用于 Med-LLMs 的关键医学 NLP 任务、数据资源和评估方法。
• 调研 Med-LLMs 在临床决策支持、报告生成和教育等领域的医疗应用。
• 讨论公平性、问责制、隐私、鲁棒性以及监管考量等方面的挑战，以促进可信度。
• 概述 Med-LLMs 在临床实践与政策情境中的潜在未来方向。

提出的方法

• 回顾 LLM 发展的里程碑及架构（Transformer、自注意力、预训练、微调）的历史进展。
• 描述医学领域的适应，包括医学语料库、知识图谱、检索增强生成，以及与人为对齐。
• 总结训练/微调技术（SFT、IFT/IPT、PEFT、RLHF、ICL）及其医学变体。
• 编目主要的 Med-LLM 任务（Med-IE、Med-QA、Med-NLI、Med-Gen）及代表性数据集。
• 讨论伦理、安全、隐私与监管方面的考量并提出未来方向。

实验结果

研究问题

• RQ1 哪些关键的技术进步使 Med-LLMs 能在医疗场景中发挥作用？
• RQ2 Med-LLMs 如何进行训练、微调以及在医疗任务和安全性方面评估？
• RQ3 哪些领域和任务最能从 Med-LLMs 受益（如临床决策支持、教育、报告生成等）？
• RQ4 Med-LLMs 面临的主要伦理、隐私、问责和鲁棒性挑战有哪些，如何应对？
• RQ5 为在医疗领域负责任地推进 Med-LLMs，需要哪些未来方向和政策考量？

主要发现

• Med-LLMs 从通用 LLMs 演变而来，通过领域特定的适应，利用医学语料库和知识图谱来提升临床推理。
• 微调策略（SFT、IFT/IPT、PEFT）以及以人为反馈的 RLHF 是使 Med-LLMs 与医疗需求和安全性保持一致的核心。
• Med-LLMs 在临床决策支持、报告生成和医学教育方面具有强大潜力，数据集涵盖多种语言和任务。
• 伦理、隐私与监管考量（GDPR、HIPAA）对于可信部署至关重要，同时还需满足鲁棒性与可解释性要求。
• 未来方向包括提升可解释性、制定政策，以及将其整合到临床工作流程中，以确保 Med-LLMs 的负责任扩展。