Phy-BP: Physics-Constrained Deep Learning for Contactless Blood Pressure Monitoring via Triaxial Bodyseismography

Forum topic · 小凯 · 2026-08-26

Summary

A 2025 arXiv paper (2508.17628) by Yuanyuan Zhang, Yida Zhang, and Jiahui Li introduces Phy-BP, a non-invasive blood pressure estimation framework based on triaxial bodyseismography (BSG), an extension of ballistocardiography (BCG). Traditional BCG signals suffer from body-bed contact variations, shifted fiducial points, and individual hemodynamic differences that degrade model generalizability. Phy-BP addresses this in two ways: first, an adaptive quality-control algorithm selects BSG segments rich in cardiogenic components by jointly considering neighboring beat patterns and universal cardiogenic templates; second, a physical model describing 3D wave propagation in the human-bed system is embedded into a deep learning model to capture the intrinsic coupling between triaxial BSG signals driven by a single cardiac excitation, aligning multi-axis features during training. Experiments on a 162-hour hospital dataset from 21 subjects show Phy-BP dynamically filters low-quality measurements and remains reliable even with limited training samples.

Overview

Field: Machine Learning Authors: Yuanyuan Zhang, Yida Zhang, Jiahui Li Published: 2025-08-26 arXiv: 2508.17628

Abstract

Ballistocardiography (BCG) is promising for unobtrusive long-term blood pressure (BP) monitoring in laboratory settings, but traditional BCG signals are vulnerable to variations in body-bed interaction, with shifted fiducial points in the temporal or amplitude axis. BP also varies with personal hemodynamic changes, causing misaligned representations that affect model generalizability and robustness.

This work proposes Phy-BP, a non-invasive BP estimation framework based on triaxial bodyseismography (BSG), an extension of BCG.

1. An adaptive quality-control algorithm selects BSG segments enriched with cardiogenic components by jointly considering neighboring beat patterns and universal cardiogenic templates. 2. A physical model describing 3D wave propagation in the human-bed system is established and embedded into a deep learning model, capturing the intrinsic coupling between triaxial BSG signals driven by a single cardiac excitation. Multi-axis features are thereby aligned during training, improving robustness against distortions in real-world scenarios.

Experiments on a 162-hour hospital dataset collected from 21 subjects demonstrate that Phy-BP can dynamically filter low-quality measurements, and the physics-consistency-constrained deep learning model still provides reliable BP monitoring when training samples are limited.

*Note: the original abstract is truncated at "a physical model is established to describe..." in the source post.*

Phy-BP: Physics-Constrained Deep Learning for Contactless Blood Pressure Monitoring via Triaxial Bodyseismography

Summary

Overview

Abstract

Tags