Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Guiding IoT-Based Healthcare Alert Systems with Large Language Models

Yulan Gao, Ziqiang Ye, Ming Xiao, Yue Xiao, Dong In Kim

TL;DR

IoT-based healthcare alert systems must balance personalized alert accuracy with strong privacy protections in resource-constrained environments. The authors introduce LLM-HAS, a framework that combines an LLM-guided Mixture-of-Experts on edge servers with diffusion-based denoising and DRL (DDPG) for health-alert decisions, augmented by conversational user feedback to improve QoE. Key contributions include a privacy-preserving mixed-content data pipeline, LLM-enabled MoE for selecting task-specific DRL experts, a reward grounded in $-$\{P$(MA)$+P$(FA)\}, and a feedback loop that continuously tunes the system; validated on the MHEALTH dataset, showing reduced false/missed alerts and robust performance under noise. This approach demonstrates a practical path toward privacy-conscious, adaptive, GAI-powered HAS suitable for real-world deployment and user-centric healthcare management.

Abstract

Healthcare alert systems (HAS) are undergoing rapid evolution, propelled by advancements in artificial intelligence (AI), Internet of Things (IoT) technologies, and increasing health consciousness. Despite significant progress, a fundamental challenge remains: balancing the accuracy of personalized health alerts with stringent privacy protection in HAS environments constrained by resources. To address this issue, we introduce a uniform framework, LLM-HAS, which incorporates Large Language Models (LLM) into HAS to significantly boost the accuracy, ensure user privacy, and enhance personalized health service, while also improving the subjective quality of experience (QoE) for users. Our innovative framework leverages a Mixture of Experts (MoE) approach, augmented with LLM, to analyze users' personalized preferences and potential health risks from additional textual job descriptions. This analysis guides the selection of specialized Deep Reinforcement Learning (DDPG) experts, tasked with making precise health alerts. Moreover, LLM-HAS can process Conversational User Feedback, which not only allows fine-tuning of DDPG but also deepen user engagement, thereby enhancing both the accuracy and personalization of health management strategies. Simulation results validate the effectiveness of the LLM-HAS framework, highlighting its potential as a groundbreaking approach for employing generative AI (GAI) to provide highly accurate and reliable alerts.

Guiding IoT-Based Healthcare Alert Systems with Large Language Models

TL;DR

IoT-based healthcare alert systems must balance personalized alert accuracy with strong privacy protections in resource-constrained environments. The authors introduce LLM-HAS, a framework that combines an LLM-guided Mixture-of-Experts on edge servers with diffusion-based denoising and DRL (DDPG) for health-alert decisions, augmented by conversational user feedback to improve QoE. Key contributions include a privacy-preserving mixed-content data pipeline, LLM-enabled MoE for selecting task-specific DRL experts, a reward grounded in \{P+P$(FA)\}, and a feedback loop that continuously tunes the system; validated on the MHEALTH dataset, showing reduced false/missed alerts and robust performance under noise. This approach demonstrates a practical path toward privacy-conscious, adaptive, GAI-powered HAS suitable for real-world deployment and user-centric healthcare management.

Abstract

Healthcare alert systems (HAS) are undergoing rapid evolution, propelled by advancements in artificial intelligence (AI), Internet of Things (IoT) technologies, and increasing health consciousness. Despite significant progress, a fundamental challenge remains: balancing the accuracy of personalized health alerts with stringent privacy protection in HAS environments constrained by resources. To address this issue, we introduce a uniform framework, LLM-HAS, which incorporates Large Language Models (LLM) into HAS to significantly boost the accuracy, ensure user privacy, and enhance personalized health service, while also improving the subjective quality of experience (QoE) for users. Our innovative framework leverages a Mixture of Experts (MoE) approach, augmented with LLM, to analyze users' personalized preferences and potential health risks from additional textual job descriptions. This analysis guides the selection of specialized Deep Reinforcement Learning (DDPG) experts, tasked with making precise health alerts. Moreover, LLM-HAS can process Conversational User Feedback, which not only allows fine-tuning of DDPG but also deepen user engagement, thereby enhancing both the accuracy and personalization of health management strategies. Simulation results validate the effectiveness of the LLM-HAS framework, highlighting its potential as a groundbreaking approach for employing generative AI (GAI) to provide highly accurate and reliable alerts.
Paper Structure (17 sections, 5 figures)

This paper contains 17 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Overview of IoT-based HAS
  3. The Proposed LLM-HAS Framework
  4. Key Research Challenges
  5. The Proposed LLM-HAS Framework
  6. LLM-Based Mixed-Content Data Creation
  7. LLM-enabled MoE Framework
  8. Health Alert Decision-Making
  9. Conversational User Feedback of LLM-HAS
  10. Experiments of LLM-HAS
  11. Experimental Setting
  12. Results and Discussion
  13. Future Directions
  14. Autonomous LLM-HAS Framework
  15. Energy-Efficient Wireless Transmission Methods
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Architecture and workflow of the proposed LLM-HAS framework: Once the system receives extra text-based descriptions of occupation or status, LLM processes the data and selects the appropriate experts required to address the users' specific needs.
  • Figure 2: Processes for generating and pre-processing mixed-content data: a case study illustration.
  • Figure 3: An example of dynamic conversational user feedback in the LLM-HAS framework.
  • Figure 4: Performance comparison of the probability of FA and MA versus different level of noise under scenario 1 and scenario 2.
  • Figure 5: Comparison of performance in terms of FA rate and MA rate, respectively.