GlucoFM is a lightweight foundation model for continuous glucose monitoring that aligns irregular CGM data to a 24‑hour grid and splits glucose dynamics into slow‑varying trend and short‑term deviation streams. Pre‑trained on over 109,000 hours of unlabeled recordings, it outperforms existing CGM‑specific models on seven phenotype‑classification tasks, improving average PR‑AUC by 4.1 points and enabling strong cross‑dataset transfer and few‑shot adaptation. When combined with meal, nutrition, and subject context, its frozen encoder delivers the lowest two‑hour postprandial glycemic response errors for trajectory, incremental AUC, peak rise, and peak timing metrics.
By Zechen Li, Keerthana Natarajan, Weizhi Zhang, Menglian Zhou, Simon A. Lee, Yuwei Zhang, Maxwell A. Xu, Zeinab Esmaeilpour, Flora D. Salim, Mark Malhotra, Lindsey Sunden, Shwetak Patel, Yuzhe Yang, Ahmed A. Metwally
arXiv:2609.28199v1 Announce Type: new
Abstract: Long physiological recordings contain many routine measurements, while predictive information is often concentrated in rare events, sustained burden, a...
By Tian Zhou, Bingqing Peng, Linxiao Yang, Wenwei Wang, Mengni Ye, Beverly Jin, Zuyi Zhu, Jinjie Gu, Liang Sun
Long physiological recordings contain many routine measurements, while predictive information is often concentrated in rare events, sustained burden, and recurring temporal patterns. Masked autoencodi...
The study evaluates time‑series foundation models for continuous glucose monitoring (CGM) forecasting across eight public datasets covering Type 1, Type 2, and non‑diabetes populations. Zero‑shot foundation models did not consistently beat strong task‑specific baselines, but lightweight fine‑tuning of models like Chronos‑Bolt improved root‑mean‑square error by up to 18% in both in‑distribution and out‑of‑distribution settings. Incorporating multimodal dietary context via CGMacros and a residual‑based fusion framework further reduced overall RMSE by ~3% and postprandial RMSE by ~15%, indicating that dietary signals add clinically meaningful value beyond CGM alone.
By Bowen Zhang, Hsiu-Wen Cheng, Hongyu Yang, Evie L. Shen, Joleen Vansomphone, Yuna Li, Kerry Zhou, Zitian Qu, Suning Zhao, Xiangning Deng, Hua Zhou, Jin J. Zhou
arXiv:2607. 21117v1 Announce Type: cross Abstract: Preprocessing blood glucose time-series data is a critical yet often overlooked step in developing data-driven methods for diabetes management, particularly for type 1 diabetes.
By Davide Marelli, Giorgia Rigamonti, Mirko Paolo Barbato, Paolo Napoletano
arXiv:2608. 01820v1 Announce Type: cross Abstract: Non-invasive blood glucose level (BGL) estimation from photoplethysmography (PPG) holds great promise for wearable health monitoring, but results across studies are hard to compare due to inconsistent datasets, data leakage, and non-standardized evaluation metrics.
By Supraja Ramesh, Markus Neufeld, Michael K\"uttner, Tobias R\"oddiger, Michael Beigl
FemWear is a parameter‑efficient wearable foundation model specifically tailored for women's health. It repurposes a pretrained multimodal wearable backbone by training only 239,236 encoder parameters—just 1.11% of the original 21.54M—using low‑rank residual adapters and causal task‑family heads to create a shared longitudinal representation for menstrual, symptom, affective, sleep/recovery, autonomic, activity, and pregnancy outcomes. Evaluations across six cohorts and 63 metrics show improvements in cycle‑phase macro‑F1 and reductions in mean absolute error for cramps, mood symptoms, and sleep problems, while maintaining the OpenMHC ability‑retention benchmark.
By Yifan Wang, Chenzhong Li
The paper evaluates how readout settings and operating points affect event-stream candidate selection in event-triggered recording. Using 520 DailyDVS‑200 clips and 47 participants, the authors compare several encoder configurations—including DINO‑CLS, patch mean, and MobileNetV3‑Small—across different accumulation durations and view counts, reporting metrics such as target capture, false triggers, and partial AUC. The study finds that lightweight baselines and specific operating points significantly influence performance, while no clear advantage is observed for the JEPA‑specific method.
By Yifan Shuai
arXiv:2609.08772v1 Announce Type: new
Abstract: Large Language Models (LLMs) are increasingly being investigated for physiological time-series prediction, yet their effectiveness may depend not only...
By Andrea Apicella, Pasquale Arpaia, Matteo Orefice, Andrea Pollastro, Roberto Prevete
The paper examines how the representation of physiological data affects the performance of large language models (LLMs) in predicting post‑meal blood glucose events for people with type 1 diabetes. Using the OhioT1DM dataset, the authors compare zero‑shot and few‑shot prompt‑based LLMs across 30, 60, and 90‑minute horizons, varying the textual encoding of glucose readings, derived descriptors, and contextual variables such as insulin, meals, carbs, and activity. Results show that while conventional supervised models excel at hyperglycemia prediction, certain prompt‑based LLM configurations outperform them for hypoglycemia, and that the way data is presented to the model is a key determinant of success, with added context not consistently improving outcomes.
Non-invasive blood glucose level (BGL) estimation from photoplethysmography (PPG) holds great promise for wearable health monitoring, but results across studies are hard to compare due to inconsistent datasets, data leakage, and non-standardized evaluation metrics. We present the first reproducible, extensible evaluation pipeline and use it to reassess five representative PPG-based BGL methods on published datasets under three increasingly strict data-split protocols: random window-level, participant-aware, and leave-some-participants-out (LSPO).
arXiv:2601. 05353v2 Announce Type: replace Abstract: Accurate blood glucose forecasting using continuous glucose monitoring (CGM) data can support the early prediction of dysglycemic risk.
By Shovito Barua Soumma, Hassan Ghasemzadeh