Evaluation of Pacemaker Battery Level Estimation using Large Language Model and Remote Monitoring Historical Data | JoVE Visualize

Area of Science:

Biomedical Engineering
Artificial Intelligence in Healthcare
Medical Device Monitoring

Background:

Remote monitoring (RM) of pacemakers is widespread, offering valuable patient data but facing challenges in manual data extraction and aggregation.
Current RM systems generate monthly reports, resulting in scarce data that hinders comprehensive patient analysis and unification.
The manual aggregation of diverse RM data formats imposes a significant workload on healthcare professionals, limiting efficient data utilization.

Purpose of the Study:

To develop an automated pacemaker battery level estimation system integrating large language models (LLMs) with historical remote monitoring (RM) data.
To address the challenges of manual data extraction, aggregation, and the scarcity of data in current RM systems for pacemakers.
To create a user-friendly system for clinicians to easily access and analyze similar patient data without requiring technical expertise.

Main Methods:

Developed a system utilizing LLMs to automate the creation and querying of a relational database from pacemaker RM data.
Employed OpenAI's generative pre-trained transformers (GPTs) to calculate high-order approximation curves from RM history for simulating pacemaker discharge characteristics.
Utilized the Akaike Information Criterion (AIC) to select the optimal approximation curve for developing the pacemaker battery remaining capacity prediction system.

Main Results:

The developed system demonstrated feasibility in integrating RM data with LLMs for pacemaker condition monitoring.
The gpt-4o-latest model achieved the lowest root mean square error (RMSE) of 0.0124mV in battery level estimation.
LLM-based data analysis facilitated easier identification of similar patient data from the relational database without technical skills.

Conclusions:

Integrating LLMs with RM data offers a feasible and efficient approach for future pacemaker condition monitoring.
The developed system reduces the workload for healthcare professionals by automating data analysis and database management.
This LLM-driven solution enhances the potential for more accurate and accessible pacemaker data analysis, improving patient care.