PaperarXiv:2509.19774
PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection
IEEE ICDM 2026 (repo tag "[ICDM 2026]"
TL;DR
Stage 2: PPG-conditioned latent rectified flow (straight-line transport noise → ECG latent), decoded to waveform. Includes a formal argument that the aligned encoder is needed for stable synthesis. Four datasets. Downstream: MC-MED multi-label CVD (ICD categories) with a fixed Net-1D classifier → macro-AUROC 0.631 (best; +19.1% vs best PPG-only baseline, +3.8% vs best generated-ECG baseline). MIMIC-AFib binary AF with VGG-13 → accuracy 0.82, F1 0.87.
- Form factor: finger/clinical PPG (MC-MED, MIMIC)
Abstract
Electrocardiography (ECG) is the clinical gold standard for cardiovascular disease (CVD) assessment, yet continuous monitoring is constrained by the need for dedicated hardware and trained personnel. Photoplethysmography (PPG) is ubiquitous in wearable devices and readily scalable, but it lacks electrophysiological specificity, limiting diagnostic reliability. While generative methods aim to translate PPG into clinically useful ECG signals, existing approaches are limited by the misalignment of physiological semantics in generative models and the complexity of modeling in high-dimensional signals. To address these limitations, we propose PPGFlowECG, a two-stage framework that aligns PPG and ECG in a shared latent space using the CardioAlign Encoder and then synthesizes ECGs with latent rectified flow. We further provide a formal analysis of this coupling, showing that the CardioAlign Encoder is necessary to guarantee stable and semantically consistent ECG synthesis under our formulation. Extensive experiments on four datasets demonstrate improved synthesis fidelity and downstream diagnostic utility. These results indicate that PPGFlowECG supports scalable, wearable-first CVD screening when standard ECG acquisition is unavailable.
Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
How it works
Cite this paper
Fang, X., Jin, J., Wang, H., Liu, C., Cai, J., Xiao, Y., Nie, G., Liu, B., Huang, S., Li, H., & Hong, S. (2025). PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection. IEEE ICDM 2026. https://arxiv.org/abs/2509.19774
Fang, Xiaocheng, Jiarui Jin, Haoyu Wang, Che Liu, Jieyi Cai, Yujie Xiao, and Guangkun Nie, et al. 2025. "PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection." IEEE ICDM 2026. https://arxiv.org/abs/2509.19774.
Fang, X., Jin, J., Wang, H., Liu, C., Cai, J., Xiao, Y. and Nie, G. et al. (2025) 'PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection', IEEE ICDM 2026. Available at: https://arxiv.org/abs/2509.19774
Fang, Xiaocheng, et al. "PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection." IEEE ICDM 2026, 2025. https://arxiv.org/abs/2509.19774.
X. Fang et al., "PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection," IEEE ICDM 2026, 2025, [Online]. Available: https://arxiv.org/abs/2509.19774.
@article{fang2025ppgflowecg,
title = {{PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection}},
author = {Fang, Xiaocheng and Jin, Jiarui and Wang, Haoyu and Liu, Che and Cai, Jieyi and Xiao, Yujie and Nie, Guangkun and Liu, Bo and Huang, Shun and Li, Hongyan and Hong, Shenda},
year = {2025},
journal = {IEEE ICDM 2026},
eprint = {2509.19774},
archivePrefix = {arXiv}
}Added to the ChatPPG research hub