ChatPPG.
Join free
Model

FairTune

Bias-aware fine-tuning of PPG-GPT for heart-rate prediction across ICU, wrist, and smartphone PPG

From “FairTune: A Bias-Aware Fine-Tuning Framework Towards Fair Heart Rate Prediction from PPG”
PPGarXiv preprintWeights: not released
Paper Cite
PreprintCatalog paper date: Sep 20, 2025Added to atlas: Oct 26, 2026Source
CitationsView DOI record (opens in a new tab)

Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.

FairTune is a bias-aware fine-tuning recipe for PPG-GPT, not a new pretrained encoder. The authors start from a PPG-GPT checkpoint pretrained on UCSF ICU data (over 200 million segments; four sizes from 19M to 1B) and fully fine-tune it for heart-rate regression. They resample PPG-DaLiA (wrist, 64 Hz), BUT-PPG (Xiaomi Mi9 camera, 30 Hz), and a MIMIC ICU set at 125 Hz to 40 Hz and use 1-second windows. The MIMIC extract is cited via Charlton et al. 2022 and is not tied to one hub dataset page. Table 2 reports unbalanced fine-tuning MAE 17.62 bpm on DaLiA, down from a pretrained MAE 89.54 bpm (80%). The same table reports MAE 19.23 bpm on BUT-PPG, down from 78.22 bpm (75%). It reports MAE 24.73 bpm on MIMIC, down from 93.81 bpm (74%). Same-dataset Table A1 unbalanced MAE is 4.28 bpm on DaLiA, 9.94 bpm on BUT-PPG, and 9.76 bpm on MIMIC. The gender gap is the absolute male–female MAE difference. Table 3 does not shrink it on every transfer: DaLiA to MIMIC moves from 0.80 bpm unbalanced to 1.06 bpm with inverse-frequency weighting, 1.20 bpm with GroupDRO, and 1.16 bpm with adversarial debiasing, while DaLiA to BUT-PPG GroupDRO lowers the gap from 2.36 to 0.68 bpm. The PDF uses a Proceedings of Machine Learning Research class header and names no volume. No code URL is given.

Signals
PPG
Authors

Panchumarthi, Kataria, Wu, Hu, Fedorov, Kwak (Emory; Wu also University of Oklahoma)

arXiv:2509.16491v1 (20 Sep 2025); DOI 10.48550/arXiv.2509.16491; PDF metadata licence CC BY 4.0; no code URL; PMLR class header, no volume

Full fine-tune of PPG-GPT. Reported results use 19M; the paper also scales from 19M to 1B. Checkpoint pretrained on UCSF ICU data, over 200M segments

Table 2 unbalanced MAE

DaLiA 89.54 to 17.62 bpm (80%), BUT-PPG 78.22 to 19.23 bpm (75%), MIMIC 93.81 to 24.73 bpm (74%). Table A1 same-dataset unbalanced MAE: 4.28, 9.94, 9.76 bpm

Table 3, 19M medians, DaLiA to BUT-PPG: unbalanced MAE 19.23 gap 2.36; IF 13.44 / 1.30; GroupDRO 15.47 / 0.68; ADV 14.49 / 0.70 bpm

Table 3 does not shrink every gap. BUT-PPG to DaLiA: unbalanced 6.99 vs IF 6.86 bpm. DaLiA to MIMIC: unbalanced 0.80 vs IF 1.06, GroupDRO 1.20, ADV 1.16 bpm. At 1B, DaLiA to MIMIC is MAE 26.66 gap 1.27 (19M: 24.74 and 0.80); MIMIC to BUT-PPG is MAE 16.23 gap 2.65 (19M: 18.68 and 4.93)

Table 1 counts samples, not subjects: DaLiA 35,099 female (54.3%) and 29,598 male (45.7%); BUT-PPG 1,986 (51.1%) and 1,902 (48.9%); MIMIC 6,804 (62.3%) and 4,123 (37.7%). Table 4 MMD male vs female: unbalanced 0.004478, IF 0.003896 (authors: 13%)

At a glance
Type
FM adaptation
Modalities
PPG
Form factor
PPG gpt bias aware finetune
Parameters
19M–1B
Architecture
Full-layer fine-tune of PPG-GPT, a decoder-only transformer with 40-sample patches. Heart-rate regression with inverse-frequency weighting, GroupDRO, or adversarial debiasing. Signals resampled to 40 Hz and cut into 1-second windows
Venue
arXiv preprint
Year
2025
arXiv
2509.16491 (opens in a new tab)
DOI
10.48550/arXiv.2509.16491 (opens in a new tab)
Code
Not released
Weights
Not released
Licence
CC BY 4.0 (opens in a new tab)
Caveats
FairTune fine-tunes PPG-GPT. It is not a new pretrained foundation model, and this preprint gives no code or weight URL. The fairness axis in the experiments is binary gender. Age, skin tone, and race are not evaluated. Table 3 medians do not show a smaller gender gap on every transfer, and some mitigation rows raise MAE. The MIMIC ICU PPG set is cited through Charlton et al. 2022. It is not linked to one hub dataset page.
Datasets
Papers
Added
Oct 26, 2026
Updated
Oct 26, 2026

Abstract

Foundation models pretrained on physiological data such as photoplethysmography (PPG) signals are increasingly used to improve heart rate (HR) prediction across diverse settings. Fine-tuning these models for local deployment is often seen as a practical and scalable strategy. However, its impact on demographic fairness—particularly under domain shifts—remains underexplored. We fine-tune PPG-GPT — a transformer-based foundation model pretrained on intensive care unit (ICU) data — across three heterogeneous datasets (ICU, wearable, smartphone) and systematically evaluate the effects on HR prediction accuracy and gender fairness. While fine-tuning substantially reduces mean absolute error (up to 80%), it can simultaneously widen fairness gaps, especially in larger models and under significant distributional characteristics shifts. To address this, we introduce FairTune, a bias-aware fine-tuning framework in which we benchmark three mitigation strategies: class weighting based on inverse group frequency (IF), Group Distributionally Robust Optimization (GroupDRO), and adversarial debiasing (ADV). We find that IF and GroupDRO significantly reduce fairness gaps without compromising accuracy, with effectiveness varying by deployment domain. Representation analyses further reveal that mitigation techniques reshape internal embeddings to reduce demographic clustering. Our findings highlight that fairness does not emerge as a natural byproduct of fine-tuning and that explicit mitigation is essential for equitable deployment of physiological foundation models.

Verbatim fromarXiv· arXiv metadata (abstracts are released under CC0)

FairTune: ChatPPG research summary card

Key results

17.62 bpm
MAE
Table 2 reports unbalanced fine-tuning MAE 17.62 bpm on DaLiA, down from a pretrained MAE 89.54 bpm (80%)
19.23 bpm
MAE
The same table reports MAE 19.23 bpm on BUT-PPG, down from 78.22 bpm (75%)
24.73 bpm
MAE
It reports MAE 24.73 bpm on MIMIC, down from 93.81 bpm (74%)

How it works

How FairTune works: Input signals → Model → Downstream tasksInput signals: PPG (Photoplethysmogram). Datasets: PPG-DaLiA, Brno University of Technology Smartphone PPG Database (BUT PPG). Model: FairTune, 19M–1B parameters (Transformer · Masked reconstruction). Downstream tasks: Heart rate.Input signalsPPG: PhotoplethysmogramPPGPhotoplethysmogramDatasets: PPG-DaLiA, Brno University of Technology Smartphone PPG Database (BUT PPG)DatasetsPPG-DaLiA, Brno University of Technology…Model×NFairTune19M–1B parametersTransformer · Masked reconstructionDownstream tasksHeart rateHeart rate

Datasets linked in the hub; pretraining vs evaluation roles follow the paper.

Pretraining data
PPG-GPT checkpoint pretrained on UCSF ICU PPG, over 200 million segments. FairTune does not run a new pretraining
Architecture
Full-layer fine-tune of PPG-GPT, a decoder-only transformer with 40-sample patches. Heart-rate regression with inverse-frequency weighting, GroupDRO, or adversarial debiasing. Signals resampled to 40 Hz and cut into 1-second windows
Inputs
PPG
Approach
PPG · foundation model · fairness · heart rate · groupdro · emory

Datasets· 2

Papers· 1

Cite

APA

Panchumarthi, L. Y., Kataria, S., Wu, Y., Hu, X., Fedorov, A., & Kwak, H. G. (2025). FairTune: A Bias-Aware Fine-Tuning Framework Towards Fair Heart Rate Prediction from PPG. arXiv preprint. https://doi.org/10.48550/arXiv.2509.16491

BibTeX

@misc{panchumarthi2025fairtune,
  title         = {{FairTune: A Bias-Aware Fine-Tuning Framework Towards Fair Heart Rate Prediction from PPG}},
  author        = {Panchumarthi, Lovely Yeswanth and Kataria, Saurabh and Wu, Yi and Hu, Xiao and Fedorov, Alex and Kwak, Hyunjung Gloria},
  year          = {2025},
  howpublished  = {arXiv preprint},
  doi           = {10.48550/arXiv.2509.16491},
  eprint        = {2509.16491},
  archivePrefix = {arXiv}
}

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

History

  1. Added to the ChatPPG research hub

Connections

Open in the graph explorer
Dataset: PPG-DaLiA (uses dataset)PPG-DaLiAPPG-DaLiADataset: Brno University of Technology Smartphone PPG Database (BUT PPG) (uses dataset)Brno University of…Brno Unive…Signal: PPG (uses signal)PPGPPGPaper: FairTune: A Bias-Aware Fine-Tuning Framework Towards Fair Heart Rate Prediction from PPG (paper and model)FairTune: A Bias-Aw…FairTune:…FairTuneFairTune
Signals 1
PPG

Same tasks, data and authors

Papers

Models

Datasets