ChatPPG.
Join free
First page or project page of “Cross-subject PPG-to-ECG reconstruction: open-data benchmark, model weights and notebook (OSF)” (opens in a new tab)

Paperdoi:10.17605/osf.io/gyd8f

Cross-subject PPG-to-ECG reconstruction: open-data benchmark, model weights and notebook (OSF)

Preprint2026Added Oct 25, 2026

  • MR

Authors: M. Jaideep Rao

OSF project 2026-07-06 (materials for a preprint)

Publisher (opens in a new tab) Cite
First page or project page of “Cross-subject PPG-to-ECG reconstruction: open-data benchmark, model weights and notebook (OSF)” (opens in a new tab)

TL;DR

OSF materials for the preprint 'Cross-Subject Reconstruction of the Electrocardiogram from Single-Site Photoplethysmography: An Open-Data Benchmark and Information-Ceiling Finding'. Includes a Colab notebook (data prep, U-Net / STFT-loss / conditional-GAN training and evaluation), trained weights (ppg2ecgvitaldb.keras) and links to public sources (BIDMC via PhysioNet; VitalDB). Reports subject-independent reconstruction Pearson r with an observed information ceiling of ~0.38.

  • Form factor: clinical_finger

Abstract

Materials for the preprint "Cross-Subject Reconstruction of the Electrocardiogram from Single-Site Photoplethysmography: An Open-Data Benchmark and Information-Ceiling Finding." Includes the Google Colab notebook (data preparation, U-Net / STFT-loss / conditional-GAN training and evaluation), the trained model weights (ppg2ecg_vitaldb.keras), and links to the public source datasets (BIDMC via PhysioNet; VitalDB). Reports subject-independent reconstruction performance (Pearson r) and the observed information ceiling of ~0.38.

Verbatim from DataCite (opens in a new tab) · Description as registered with DataCite

Results on ChatPPG leaderboards

All leaderboards

No results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.

Tasks & signals

Tasks

Signal denoising

Signals

PPGECG

Keywords

PPG-to-ECGreconstructionU-NetcGANbenchmarkBIDMCVitalDBsubject-independent

How it works

How Method works: Setup → Method → TargetSetup: Signals (PPG, ECG); Form factor (clinical_finger). Method: Method (Convolutional network · Masked reconstruction). Target: Signal denoising.SetupSignals: PPG, ECGSignalsPPG, ECGForm factor: clinical_fingerForm factorclinical_fingerMethodMethodConvolutional network · Masked…TargetSignal denoisingSignal denoising

Cite this paper

APA

Rao, M. J. (2026). Cross-subject PPG-to-ECG reconstruction: open-data benchmark, model weights and notebook (OSF). OSF project 2026-07-06. https://doi.org/10.17605/osf.io/gyd8f

BibTeX

@article{rao2026crosssubject,
  title         = {{Cross-subject PPG-to-ECG reconstruction: open-data benchmark, model weights and notebook (OSF)}},
  author        = {Rao, M. Jaideep},
  year          = {2026},
  journal       = {OSF project 2026-07-06},
  doi           = {10.17605/osf.io/gyd8f}
}

Built from the hub record. Check it against the publisher before submitting.

Added to the ChatPPG research hub

Same tasks, data and authors

Papers

Models

Datasets