2024 World Conference on Lung Cancer (WCLC) - ePosters-EP.17C.02 Nodule-AFM: Construction and Application of LLM and Multi-Agent Collaborative Based Active Follow-Up Management for Pulmonary Nodule Patients

EP.17C.02 Nodule-AFM: Construction and Application of LLM and Multi-Agent Collaborative Based Active Follow-Up Management for Pulmonary Nodule Patients

Back to course

Pdf Summary

The document details the development and implementation of Nodule-AFM, a novel follow-up management system leveraged by large language models (LLMs) and a multi-agent collaborative framework for patients with pulmonary nodules. This system addresses inefficiencies in the current predominantly manual follow-up procedures that are crucial for early lung cancer detection and intervention.<br /><br />Nodule-AFM applies artificial intelligence (AI) and machine learning (ML) to enhance accuracy and streamline the follow-up process. It utilizes an open-source LLM fine-tuned with approximately 1000 follow-up question-and-answer samples and operates within a system architecture composed of three main agents: the Database Agent, Active Q&A Agent, and Summarization Agent. These agents collectively manage patient data retrieval, initiate follow-up inquiries, and summarize interaction results, respectively, all executed on a lightweight browser/server platform created using Langchain and Streamlit.<br /><br />The system was evaluated using a dataset of 100 cases, delivering a high completion rate and substantial performance in information extraction and summarization, with a 94% success rate in conversations, 84% in extracting dialogue information, and 79% in summarizing dialogue accurately. These figures indicate a significant improvement over manual follow-up methods.<br /><br />The discussion emphasizes the potential of AI-driven enhancements in follow-up management, showcasing improved efficiency and accuracy compared to traditional methods. Future work aims to optimize the model further, incorporate more personalized patient-specific variables, expand applications to other medical conditions, and integrate feedback mechanisms for continuous improvement.<br /><br />The Nodule-AFM system signifies a progressive step in the medical field by automating follow-up tasks, which can reduce healthcare providers' workloads, providing a more efficient, accurate, and personalized approach to patient management.

Asset Subtitle

Dahai Liu

Meta Tag

Speaker Dahai Liu

Topic Global Health, Health Services, and Health Economics

Keywords

Nodule-AFM

pulmonary nodules

large language models

multi-agent framework

artificial intelligence

machine learning

Langchain

Streamlit

follow-up management

healthcare automation