Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PENGUIN: General Vital Sign Reconstruction from PPG with Flow Matching State Space Model

Shuntaro Suzuki, Shuitsu Koyama, Shinnosuke Hirano, Shunya Nagashima

TL;DR

PENGUIN tackles the challenge of reconstructing multiple vital-sign waveforms from PPG with high fidelity and generalizability. It combines Optimal Transport Conditional Flow Matching with a dual-stream Flow-SSM to enable fine-grained, per-timestep conditioning on PPG, yielding continuous ECG, respiratory, and ABP waveforms from six real-world datasets. The approach outperforms task-specific and generalist baselines across multiple metrics, and ablations confirm the importance of conditioning mechanisms for morphology fidelity. This framework advances PPG-based monitoring by delivering consistent, waveform-level reconstructions suitable for diverse clinical and consumer-health applications. The use of $p_0= ext{N}(0,1)$, OT-CFM shortest-transport flow, and $25$-step ODE sampling underpins efficient, high-quality generation of vital-sign signals from PPG.

Abstract

Photoplethysmography (PPG) plays a crucial role in continuous cardiovascular health monitoring as a non-invasive and cost-effective modality. However, PPG signals are susceptible to motion artifacts and noise, making accurate estimation of vital signs such as arterial blood pressure (ABP) challenging. Existing estimation methods are often restricted to a single-task or environment, limiting their generalizability across diverse PPG decoding scenarios. Moreover, recent general-purpose approaches typically rely on predictions over multi-second intervals, discarding the morphological characteristics of vital signs. To address these challenges, we propose PENGUIN, a generative flow-matching framework that extends deep state space models, enabling fine-grained conditioning on PPG for reconstructing multiple vital signs as continuous waveforms. We evaluate PENGUIN using six real-world PPG datasets across three distinct vital sign reconstruction tasks (electrocardiogram reconstruction, respiratory monitoring, and ABP monitoring). Our method consistently outperformed both task-specific and general-purpose baselines, demonstrating PENGUIN as a general framework for robust vital sign reconstruction from PPG.

PENGUIN: General Vital Sign Reconstruction from PPG with Flow Matching State Space Model

TL;DR

PENGUIN tackles the challenge of reconstructing multiple vital-sign waveforms from PPG with high fidelity and generalizability. It combines Optimal Transport Conditional Flow Matching with a dual-stream Flow-SSM to enable fine-grained, per-timestep conditioning on PPG, yielding continuous ECG, respiratory, and ABP waveforms from six real-world datasets. The approach outperforms task-specific and generalist baselines across multiple metrics, and ablations confirm the importance of conditioning mechanisms for morphology fidelity. This framework advances PPG-based monitoring by delivering consistent, waveform-level reconstructions suitable for diverse clinical and consumer-health applications. The use of , OT-CFM shortest-transport flow, and -step ODE sampling underpins efficient, high-quality generation of vital-sign signals from PPG.

Abstract

Photoplethysmography (PPG) plays a crucial role in continuous cardiovascular health monitoring as a non-invasive and cost-effective modality. However, PPG signals are susceptible to motion artifacts and noise, making accurate estimation of vital signs such as arterial blood pressure (ABP) challenging. Existing estimation methods are often restricted to a single-task or environment, limiting their generalizability across diverse PPG decoding scenarios. Moreover, recent general-purpose approaches typically rely on predictions over multi-second intervals, discarding the morphological characteristics of vital signs. To address these challenges, we propose PENGUIN, a generative flow-matching framework that extends deep state space models, enabling fine-grained conditioning on PPG for reconstructing multiple vital signs as continuous waveforms. We evaluate PENGUIN using six real-world PPG datasets across three distinct vital sign reconstruction tasks (electrocardiogram reconstruction, respiratory monitoring, and ABP monitoring). Our method consistently outperformed both task-specific and general-purpose baselines, demonstrating PENGUIN as a general framework for robust vital sign reconstruction from PPG.
Paper Structure (12 sections, 4 equations, 2 figures, 2 tables)

This paper contains 12 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Proposed Method
  4. Flow Matching Formulation
  5. Deep State Space Model Formulation
  6. Model Architecture
  7. Experiments
  8. Datasets and Data Pre-processing
  9. Evaluation Metrics
  10. Implementation Details
  11. Results and Discussion
  12. Conclusion

Figures (2)

  • Figure 1: Framework of PENGUIN. (a) A stack of Flow-SSM blocks outputs the derivative of the vital sign flow, conditioned on PPG. (b) S5 layer S5, a variant of SSMs, is extended for sequence modeling. (c) PENGUIN is built upon the flow matching framework OT-CFM.
  • Figure 2: Qualitative comparison of reconstructed vital signs from PPG over a 4-second segment.