Orexin System
Open App Contact
A research instrument

Predicting wake instability. Before you feel it.

A wearable system that models the latent state governing wakefulness stability — what the orexin system does functionally — by fusing its observable downstream effects across pupillometry, autonomic, thermal, respiratory, and chemical channels from three coordinated body-worn devices.

7
Sensing modalities
3
Body-worn devices
1
Prediction
7 orthogonal sensing modalities → 40+ derived features → 1 prediction
1 — Eye imaging
What it provides +
The fast dynamics layer. Sub-second arousal state changes. The only sensing modality with a direct causal link to orexin neurons (Grujic et al., Nature Neuroscience, 2023). Derived features: pupil diameter, pupillary unrest index, blink rate, blink duration, PERCLOS, eyelid aperture. Sensor: OV7251 IR camera on glasses with 940nm illumination. Answers: "Is the state changing right now?"
2 — Electrodermal activity
What it provides +
Sympathetic nervous system tone — independent of cardiac signals. Orexin neurons project directly to the sympathetic outflow that controls sweat glands. A drop in EDA with a simultaneous rise in HRV signals autonomic dissociation — the systems are decoupling, which is a hallmark of orexin failure. Derived features: tonic level, phasic burst count, phasic amplitude. Sensor: Stainless-steel electrodes on glasses nose bridge. Answers: "Is the sympathetic system holding?"
3 — Cardiac (PPG + ECG)
What it provides +
Autonomic balance — the parasympathetic vs sympathetic state. Measured from two locations: wrist PPG (watch, 24/7 including sleep) and clinical-grade ECG (shirt, higher precision R-R intervals). In-ear PPG (glasses) provides a third, motion-resistant source. Derived features: heart rate, HRV (RMSSD, SDNN, LF/HF), SpO₂, ECG-derived respiration, sleep stages. Sensors: MAX30102 (watch + glasses), AD8232 (shirt). Answers: "Is the autonomic balance stable or shifting toward sleep?"
4 — Temperature gradient
What it provides +
Thermoregulatory state and circadian phase. Three temperature points: ear canal (core, glasses), wrist (distal, watch), chest (proximal, shirt). The distal-proximal gradient (DPG) — the difference between wrist and core temperature — is one of the strongest published predictors of imminent sleep onset. When the gradient narrows, the body is dumping heat to prepare for sleep. Derived features: absolute temperatures, trends, DPG, DPG rate of change. Sensors: MAX30205 ×3 locations. Answers: "Is the body preparing for a state transition?"
5 — Motion
What it provides +
Behavioral state — the essential disambiguation context. Without motion data, the model can't distinguish "heart rate is elevated because of a crash" from "heart rate is elevated because you're walking upstairs." Derived features: step count, activity classification, sleep/wake detection, micro-tremor index (atonia leak proxy). Sensor: LIS3DH accelerometer on watch. Answers: "What is the body doing right now, and should we trust the other signals?"
6 — Ambient light
What it provides +
Environmental calibration for the pupil signal. Pupil size responds to both light and autonomic state. Without subtracting the pupillary light reflex, every pupil measurement is contaminated by room brightness. The light sensor at the eye makes the difference between measuring the room and measuring the brain. Also provides cumulative light exposure history — a proxy for circadian input. Sensor: BH1750 on glasses bridge. Answers: "Is the pupil changing because of light or because of the brain?"
7 — Breath chemistry
What it provides +
Metabolic state from volatile organic compounds in breath. The chemical composition of exhaled air changes with fatigue, stress, and circadian phase. Humans can literally smell tiredness on others (Stockholm University, 2020). A gas sensor on the nose bridge — centimeters from each exhale — quantifies what the human nose detects intuitively. Trained per-user to recognize individual fatigue chemical fingerprints. Sensor: BME688 VOC gas sensor on glasses bridge. Answers: "What does the body's chemistry say about fatigue?"
3 coordinated devices, time-synchronized to <10ms
Glasses
Arousal node
Inward-facing IR camera for controlled-light pupillometry and eyelid tracking (PERCLOS). Nose-bridge electrodermal activity and VOC gas sensor for breath chemistry analysis. In-ear extension for core temperature and motion-resistant PPG. Worn during waking hours.
Watch
Metabolic node
24/7 cardiovascular and metabolic baseline. The only device worn during sleep. Tracks distal skin temperature for the distal-proximal gradient — one of the strongest predictors of imminent sleep onset.
Shirt Pod
Stability node
Clinical-grade single-lead ECG via conductive textile electrodes. ECG-derived respiration. Master clock for inter-device synchronization. High-resolution local data logging.
The Prediction
P(crash) next 20 min 12% Systems synchronized

One prediction, updated every 30 seconds: the probability of wakefulness collapse within the next 20 minutes — derived from the divergence across physiological subsystems. When all systems agree, the probability is low. When they diverge, it rises. Not a wellness score. A testable prediction with defined sensitivity, specificity, and validation endpoints.

Calibrated per individual over 2–3 weeks of passive data collection. The system learns your specific divergence pattern before crashes — not population averages. Validated against PVT reaction time, self-labeled events, and (in clinical settings) MSLT and PSG.

deep sleep drowsy calm alert active alert peak arousal
01 — Problem

No way to see wakefulness stability in real time.

Orexin (hypocretin) governs wakefulness stability — the brain's resistance to unwanted state transitions, attention collapse, and microsleep. When orexin fails, the boundary between wake and sleep becomes unstable. This is narcolepsy, but subtler dysregulation affects idiopathic hypersomnia, shift work, circadian disruption, and a growing list of neuropsychiatric conditions.

Direct orexin measurement requires a lumbar puncture — invasive, single-snapshot, confined to clinical settings. Like measuring annual income to understand daily cash flow. There is no continuous tool for wakefulness stability in free-living conditions, which limits research, blocks longitudinal studies, and leaves the rapidly expanding pipeline of orexin-targeting therapeutics without a real-time endpoint.

02 — Approach

Detecting when the body stops behaving like a synchronized system.

The orexin system doesn't just regulate wakefulness — it synchronizes the subsystems that maintain it. Pupils, heart rhythm, skin conductance, temperature regulation, and breathing patterns all march in lockstep when orexin is active. When it fails, they decouple. The decoupling is the signal.

We don't look for any single system going "bad." We look for systems disagreeing with each other:

Divergence detection — each sensing modality produces its own estimate of the body's readiness state. When all estimates agree (pupils stable, heart balanced, temperature regulated, EDA steady), the system is coherent and wakefulness is secure. When estimates diverge (pupils oscillating while heart rate is calm, temperature dumping while EDA flatlines), the physiological subsystems have lost their common coordinator. That divergence — not any single metric — is the precursor to collapse.

Seven orthogonal sensing modalities from three coordinated wearable devices, time-synchronized to within 10 milliseconds, each contributing a probabilistic estimate to multiple latent states. The variance across those estimates becomes a single prediction: the probability of wakefulness collapse within the next 20 minutes.

"Pupillometry is a reliable readout of orexin neuron activity."

Grujic et al. — Control and coding of pupil size by hypothalamic orexin neurons. Nature Neuroscience, 2023.

Most wearable systems average their signals. We measure the disagreement between them. That's much harder to replicate — and much closer to what the orexin system actually does.

03 — System

Four components. One signal.

Each device contributes a distinct subset of the autonomic signature. The system degrades gracefully — partial sensor availability still produces a confidence-weighted index.

01 / Glasses
Arousal node
Pupil · PERCLOS · EDA · Breath chemistry · Core temperature · In-ear PPG
Inward-facing IR camera with 940nm illumination for controlled-light pupillometry and eyelid tracking (PERCLOS). Nose-bridge electrodermal activity and breath chemistry sensing via VOC gas sensor. In-ear extension for core body temperature and motion-resistant heart rate.
02 / Watch
Metabolic node
HR · HRV · SpO₂ · Sleep · Wrist temperature
The always-on synchronization hub. All sensor data from glasses, shirt, and watch is time-stamped, fused, and coordinated through the watch. Every signal is raw-accessible and logged at research-grade resolution. The only device worn during sleep.
03 / Shirt Pod
Stability node
ECG · HRV · Respiration · Chest temperature
Clinical-grade single-lead ECG via conductive textile electrodes. ECG-derived respiration. Chest skin temperature. Local high-resolution logging to microSD.
04 / Pod
Power & compute
Charging · Sync · Future expansion
Power infrastructure with a forward-compatible internal architecture. Reserved expansion slots for on-device AI inference and future biofluid sensing modules as those technologies mature.
04 — Honest framing

A divergence detector. Not a molecule sensor.

This system does not measure orexin molecules. No wearable can. It detects impending failures in wakefulness by measuring divergence across physiological subsystems that are normally synchronized by the orexin system. The distinction matters: we are not estimating a peptide concentration. We are detecting when the body's subsystems stop agreeing with each other — which is what happens when their common coordinator fails.

It is a research instrument and personal data logger. It is not an FDA-cleared medical device. It is not a diagnostic tool for narcolepsy or any other condition. It should never be used for safety-critical decisions.

Status — Pre-prototype

Hardware design finalized. First build in progress.

We are actively seeking research partnerships with sleep laboratories and pharmaceutical research groups to validate wake-stability modeling against established endpoints (MSLT, PVT, PSG). Co-design input on the v1 specification is open through Q3 2026.

Get in touch →
05 — Contact
Orexin System Corp
feng@orexinsystem.com
West Vancouver, BC, Canada