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.
CHARM is a causal temporal convolutional network over 90 days of nightly ring-wearable summaries: skin-temperature delta, mean sleep HR, sleep respiratory rate, overnight HRV and lowest sleep HR (5 daily channels plus observation masks). It is pretrained on 38,064 users and 8.3 M days collected through the Terra API. The training objective combines masked-channel reconstruction with a luteal-vs-follicular phase classification loss, and the phase loss is the paper's main contribution.
Signals
Resp, Temp, HR, HRV
Form factor
Ring
Year
2026
Architecture
Causal TCN over a 90-day window of daily nocturnal features (10 input channels = 5 nightly channels (skin-temperature delta, mean sleep HR, sleep respiratory rate, overnight HRV, lowest sleep HR) + 5 observation masks); 1×1 conv to 128 ch, 4 residual dilated blocks (dilation 1/2/4/8, receptive field 61 d), 128-d per-day embedding
Pretraining data
Masked-channel reconstruction (whole channel hidden with p=0.3; loss weight 0.5) + luteal/follicular phase BCE on 38,064 users / 8.3 M days from ring wearables via Terra API (private)
License
Paper CC BY 4.0 (Research Square preprint); code proprietary, available on request; no weights
The menstrual cycle can be tracked through physiological signals captured by consumer wearables overnight — skin temperature, breathing rate, heart rate and heart rate variability — which act as proxies for the underlying hormonal variations. Other reproductive states, such as pregnancy or menopause, are often studied in isolation rather than as perturbations of the normal menstrual hormone rhythm, even though they are all governed by the same endocrine system. Here, we present CHARM (Cyclic Hormone-Aware Representation of the Menstrual cycle), a framework that learns menstrual cycle representations from wearable data by combining two training objectives: reconstruction of masked sensor channels, and a menstrual cycle phase (luteal or follicular) supervised loss. Phase supervision improves recovery of the underlying hormone dynamics from the learned representation. Without further training, frozen CHARM embeddings also generalise to multiple reproductive states and symptoms not seen during model training. The phase objective substantially improves this generalisation and is highly label-efficient, matching a strong baseline while using around 5-fold fewer labels for perimenopause and outperforming the baseline on menopause across all data regimes.
Verbatim fromthe publisher's DOI record· Publisher licence: https://creativecommons.org/licenses/by/4.0/
Datasets linked in the hub; pretraining vs evaluation roles follow the paper.
Pretraining data
Masked-channel reconstruction (whole channel hidden with p=0.3; loss weight 0.5) + luteal/follicular phase BCE on 38,064 users / 8.3 M days from ring wearables via Terra API (private)
Architecture
Causal TCN over a 90-day window of daily nocturnal features (10 input channels = 5 nightly channels (skin-temperature delta, mean sleep HR, sleep respiratory rate, overnight HRV, lowest sleep HR) + 5 observation masks); 1×1 conv to 128 ch, 4 residual dilated blocks (dilation 1/2/4/8, receptive field 61 d), 128-d per-day embedding
Inputs
Resp, Temp, HR, HRV
Approach
women's health · menstrual cycle · ovulation · pregnancy · TCN · ring
Cite the publication recorded for this model. Check it against the publisher before submitting.
Díaz, R. M., Akhtar, F., Crawford, C., Brownlee, A., & Eleftheriou, K. (2026). CHARM: cycle-phase supervision for a wearable foundation model of women's health (Terra API; Research Square 2026). Research Square preprint rs-10634632 v1 , CC BY 4.0. https://doi.org/10.21203/rs.3.rs-10634632/v1
Díaz, Rocío Mexía, Faraaz Akhtar, Cameron Crawford, Alistair Brownlee, and Kyriakos Eleftheriou. 2026. "CHARM: cycle-phase supervision for a wearable foundation model of women's health (Terra API; Research Square 2026)." Research Square preprint rs-10634632 v1 , CC BY 4.0. https://doi.org/10.21203/rs.3.rs-10634632/v1.
Díaz, R. M., Akhtar, F., Crawford, C., Brownlee, A. and Eleftheriou, K. (2026) 'CHARM: cycle-phase supervision for a wearable foundation model of women's health (Terra API; Research Square 2026)', Research Square preprint rs-10634632 v1 , CC BY 4.0. doi: 10.21203/rs.3.rs-10634632/v1.
Díaz, Rocío Mexía, et al. "CHARM: cycle-phase supervision for a wearable foundation model of women's health (Terra API; Research Square 2026)." Research Square preprint rs-10634632 v1 , CC BY 4.0, 2026. https://doi.org/10.21203/rs.3.rs-10634632/v1.
R. M. Díaz, F. Akhtar, C. Crawford, A. Brownlee, and K. Eleftheriou, "CHARM: cycle-phase supervision for a wearable foundation model of women's health (Terra API; Research Square 2026)," Research Square preprint rs-10634632 v1 , CC BY 4.0, 2026, doi: 10.21203/rs.3.rs-10634632/v1.
@misc{diaz2026charm,
title = {{CHARM: cycle-phase supervision for a wearable foundation model of women's health (Terra API; Research Square 2026)}},
author = {Díaz, Rocío Mexía and Akhtar, Faraaz and Crawford, Cameron and Brownlee, Alistair and Eleftheriou, Kyriakos},
year = {2026},
howpublished = {Research Square preprint rs-10634632 v1 , CC BY 4.0},
doi = {10.21203/rs.3.rs-10634632/v1}
}
History
Added to the ChatPPG research hub
Newsletter
Get new PPG datasets, papers and models
Get alerts when CHARM is benchmarked on new datasets, plus the weekly digest.
New datasets, papers and models
Practical guides from the hub
Unsubscribe anytime
The waveform is an illustration of a PPG pulse, not a recording.
