Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals

TL;DR

The paper addresses cuffless BP estimation from wearable biosignals by leveraging large language models (LLMs) through context-enhanced prompts and instruction tuning. It extracts 31 physiologic features from ECG/PPG signals, encodes BP-domain knowledge and user information in prompts, and fine-tunes ten open-source LLMs, comparing them against traditional baselines on the CAS-BP dataset. Results indicate that instruction-tuned LLMs can surpass conventional methods, with ablations showing meaningful gains from including domain knowledge and user context, and data-efficient tuning achievable with as little as 30% of the data. The work introduces the CBPM-LLaMA approach and demonstrates the potential of LLMs to enable accurate, cuffless BP estimation from wearable biosignals, advancing continuous cardiovascular monitoring.

Abstract

Large language models (LLMs) have captured significant interest from both academia and industry due to their impressive performance across various textual tasks. However, the potential of LLMs to analyze physiological time-series data remains an emerging research field. Particularly, there is a notable gap in the utilization of LLMs for analyzing wearable biosignals to achieve cuffless blood pressure (BP) measurement, which is critical for the management of cardiovascular diseases. This paper presents the first work to explore the capacity of LLMs to perform cuffless BP estimation based on wearable biosignals. We extracted physiological features from electrocardiogram (ECG) and photoplethysmogram (PPG) signals and designed context-enhanced prompts by combining these features with BP domain knowledge and user information. Subsequently, we adapted LLMs to BP estimation tasks through fine-tuning. To evaluate the proposed approach, we conducted assessments of ten advanced LLMs using a comprehensive public dataset of wearable biosignals from 1,272 participants. The experimental results demonstrate that the optimally fine-tuned LLM significantly surpasses conventional task-specific baselines, achieving an estimation error of 0.00 $\pm$ 9.25 mmHg for systolic BP and 1.29 $\pm$ 6.37 mmHg for diastolic BP. Notably, the ablation studies highlight the benefits of our context enhancement strategy, leading to an 8.9% reduction in mean absolute error for systolic BP estimation. This paper pioneers the exploration of LLMs for cuffless BP measurement, providing a potential solution to enhance the accuracy of cuffless BP measurement.

Figures (7)

• Figure 1: Comparison of cuff-based (a) and cuffless (b) BP measurement approaches. The cuff-based method involves the use of an inflatable cuff to occlude a peripheral artery, which can be uncomfortable. In contrast, the cuffless method calculates BP indirectly from biosignals measured by wearable sensors, avoiding the need for an inflatable cuff. This provides the advantages of comfort and unobtrusive.
• Figure 2: Framework of cuffless BP estimation using LLMs.
• Figure 3: Distribution of SBP and DBP in the CAS-BP dataset.
• Figure 4: Density Bland–Altman plots of the fine-tuned LLaMA3-8B model for SBP (a) and DBP (b) estimations. Density correlation plots of the fine-tuned LLaMA3-8B model for SBP (c) and DBP (d) estimations.
• Figure 5: BP estimation of LLaMA-3-8B with different training sizes. The solid lines represent the performance of the fine-tuned LLaMA-3-8B model, while the dashed lines indicate the performance achieved by training AdaBoost with 80% of the original data, serving as the baseline.