PulseOx-FM: A Foundation Model for Pulse Oximetry Sleep Signals
segal_eran · x · 2026-07-03
The author released a preprint introducing PulseOx-FM, a foundation model trained on approximately 7 million overnight pulse oximetry signals from over 10,000 participants in the Human Phenotype Project.
Research indicates that wearable sleep signals carry broader physiological information than standard heart rate, blood oxygen, or apnea features. PulseOx-FM outperforms engineered PPG and proprietary feature sets in age prediction, extrapolates well to external surgical cohorts, and performs strongly across 64 cardiometabolic, sleep, imaging, medication, and neuropsychiatric phenotypic targets.
It can also capture the two-year risk of developing hypertension from baseline data, along with next-day metrics like blood glucose, diet, and activity.
More from Research
- Deep Learning Course Adds Mechanistic Interpretability Lecture This Winter — CSProfKGD · 2026-10-26
- INCLUDE-2.0 benchmark evaluates regional knowledge across 91 languages and 86 countries — ABosselut · 2026-10-26
- CoRL paper: flow-matching VLAs can run 1-step Euler inference with no retraining — kaixhin · 2026-10-26
- CoVer at EMNLP 2026: Training Code LLMs to Generate and Self-Verify Simultaneously — algo_diver · 2026-10-26
- Robotic dexterity should be judged on speed, force control and tool use, not copying humans — DJiafei · 2026-10-26
- Turing-complete malware can run in PyTorch's backward pass via a zero-valued loss — retr0jirachi · 2026-10-26