Definition, measured object, and technology boundaries
An infrared thermal imaging sleep monitoring device passively receives radiation emitted by the person, bedding, and environment with a thermal-infrared focal-plane array. It forms a spatial thermogram or radiometric dataset and then analyzes nighttime posture, body movement, respiration-related change, or surface-temperature trends. Within this encyclopedia it belongs to WSIC 1.2.1, Vital-sign Monitoring Devices. Its direct observation is radiation from surfaces within the field of view, and it does not require visible illumination of the room.
Thermal imaging is not the same as near-infrared night vision. Near-infrared night vision usually illuminates a scene and records reflected radiation, whereas a thermal camera receives radiation emitted by objects. A spot or scanning non-contact infrared thermometer does not provide a complete two-dimensional time series. A passive infrared motion detector reports only zonal change or a trigger, and an infrared-emitting depth camera mainly reports distance. The word “infrared” in all of these names does not make their spectral bands, pixel meanings, or evidence interchangeable.
Documentation should also distinguish non-radiometric imagery that shows relative warm and cool patterns from a radiometric system that reports calibrated apparent temperature. Recording capability in a general thermal camera does not establish sleep-analysis capability. Bed presence, respiration, sleep stages, or abnormality alerts additionally require bed-region calibration, target tracking, quality control, algorithms, and appropriate performance evidence. Thermography used as a channel for human observation should be evaluated separately from automatic medical interpretation.
Radiometric imaging, calibration, and the data chain
A thermal camera comprises infrared optics, a detector array, readout electronics, temperature compensation, calibration parameters, a clock, processor, storage, and a network interface. Relevant specifications include spectral band, spatial resolution, field of view, instantaneous field of view, frame rate, thermal sensitivity, measurement range, and absolute temperature error. Thermal sensitivity describes the ability to distinguish small temperature differences; it is not the same as absolute accuracy at every pixel. High array resolution also does not ensure that a distant nose and mouth occupy enough pixels.
Radiometric temperature measurement converts detector signal into apparent surface temperature using calibration, target emissivity, reflected environmental temperature, distance, and atmospheric conditions. Skin generally has high emissivity, but sweat, hair, clothing, masks, viewing angle, and covering change the visible surface and the sources of error. Sheets, metal rails, mirrors, and other low-emissivity materials may reflect nearby heat sources, so a “hot spot” in an image does not necessarily originate from heat generated by that object.
Internal temperature drift, defective pixels, non-uniformity correction, automatic gain or palette remapping, compression, and lost frames can all change thermal video. False color is only a mapping from pixel values to a display scale; when the upper and lower limits change automatically, the same color need not mean the same temperature. A screenshot may also omit original radiometric values. The system should retain raw or traceable values, range, emissivity and environmental settings, correction times, timestamps, effective frame rate, and algorithm version and should mark correction interruptions or inadequate quality.
Respiration, posture, temperature, and sleep inference
Thermography can track cyclic temperature change around the nostrils or mouth as cooler room air is inhaled and warmer air is exhaled. It can alternatively analyze periodic motion of the chest, abdomen, shoulders, or bedding. These approaches are respectively closer to a thermal trace of airflow and visible-surface movement, but both depend on locating and continuously tracking a region. Head turning, prone posture, mouth breathing, a mask, an airflow device, a heavy duvet, room drafts, and large movement can move or hide the target.
Body outline and thermal distribution can also support empty-bed and occupancy detection, large movement, turns, bed exit, and coarse posture. A thermogram shows a temperature-related pattern on the exterior of skin, clothing, or bedding and cannot see through a cover. Heat patterns on a duvet have conduction delay and are not the current posture of the covered body. Skin-surface temperature is influenced by thermoregulation, sweating, airflow, contact pressure, and the environment and is not a direct measure of core temperature, blood-flow intensity, or infection.
Respiratory rate, individual breaths, pause events, sleep-wake state, and sleep stages are downstream computations from thermal video. A thermal camera does not directly measure electroencephalography, electro-oculography, electromyography, electrocardiography, respiratory effort, or arterial oxygen saturation. Nor is a nostril temperature change a calibrated volume of airflow. A claim to classify apnoea type, hypoxaemia, or sleep stage requires event-by-event or epoch-by-epoch testing against the relevant reference channels in the intended population; short-period respiratory-rate agreement cannot establish diagnostic performance.
Installation, environment, occlusion, and multi-person failure
Camera location, distance, angle, focus, and field of view should be selected from the region of interest, bed width, occupancy, common postures, and adjustable-bed positions. The nose, mouth, chest, or abdomen should still occupy enough pixels in the least favorable intended posture, while unnecessary areas are excluded. Acceptance in the actual room should include an empty-bed baseline, entry, principal postures, turning, sitting, leaving, and returning. Moving the camera, bed, or digital crop requires the scale, bed region, and target tracking to be checked again.
Changes in room temperature or humidity, air-conditioning and fan drafts, radiators, space heaters, electric blankets, hot-water bottles, windows, lamps, electronics, pets, and carers can alter the thermal background. Duvets, pillows, hair, clothing, nets, curtains, and medical masks can block skin or the airflow region. A protective window, contamination, and condensation can reduce infrared transmission. The system should be allowed an adequate stabilization period and the environment recorded, rather than interpreting a room- or bedding-driven thermal change as physiology.
With two people in one bed, overlapping outlines, fused warm regions, and identity switches can assign respiration or bed-exit events to the wrong person. A model evaluated for one occupant cannot become a multi-person system merely by widening the field. The system should also detect an out-of-frame target, lost region, inadequate thermal contrast, saturation, calibration pause, frozen frame, clock drift, full storage, power loss, network loss, and cloud outage. For alerts, documentation needs to state whether local analysis continues, how invalid data appear, who reviews them, and whether a technical failure can be mistaken for a normal state.
Performance evaluation, metrics, and procurement acceptance
Imaging evaluation should separately cover spatial resolution, target pixel count, frame rate, thermal sensitivity, uniformity, drift, calibration status, and absolute temperature error under specified conditions. An attractive false-color image is not a numerical test. Continuous outputs such as respiratory rate need valid coverage, bias, mean or median absolute error, limits of agreement, and recovery time after movement. Individual-breath or pause events additionally need sensitivity, positive predictive value, false alarms per hour or night, event-timing error, and an uninterpretable-data proportion.
The reference method must match the claim. Respiratory motion may be compared with thoracic and abdominal belts, while airflow or pause decisions require suitable nasal-pressure, thermal, or other sleep-respiratory channels. A core-temperature claim requires an appropriate temperature reference, and sleep staging requires synchronized polysomnography. A report should describe participants and nights, disease spectrum, posture, bedding, masks, camera placement, environment, reference annotations, synchronization, exclusions, and hardware and algorithm versions. Selecting only motionless, event-free, face-visible intervals overstates whole-night performance.
Before procurement, a multi-night parallel pilot in the intended bedroom should define acceptable environmental range, valid coverage, event metrics, temperature error, delay, invalid nights, privacy, and service interruption in advance and review every missing interval. Evidence must correspond to the complete product, installed configuration, and enabled functions. A detector specification, short experiment, vendor dataset, or earlier model cannot replace product acceptance. Changes to the lens, protective window, palette, calibration, codec, installation, or model should be versioned and followed by risk-proportionate regression testing.
Thermal-image privacy, network security, and physical safety
A thermogram normally lacks the colors and textures of an ordinary photograph, but it can still show body outline, posture, exposure, routine, cohabitation, care activity, and room layout. Temperature trends and event records may also be health-related data. Before activation, everyone likely to enter the field should be told what is collected, whether visible video or audio is combined, where processing occurs, who has access, how long data are retained, and how deletion works. Controls should support pausing, covering the lens, export, and audit.
Low resolution, a silhouette, cropping, edge-region data, or event-only retention can reduce circulation of raw thermograms but does not automatically make the data anonymous. Time, room, account, and behavioral patterns may still identify a person. A connected system needs unique device identity and credentials, least privilege, protection in transit and at rest, access logs, controlled sharing, signed updates, and vulnerability response. It should explain retention by mobile client, cloud, third party, and backups. Disposal, transfer, and end of service require access revocation and secure deletion.
A passive thermal camera ordinarily does not illuminate the person with infrared radiation, but visible light, near-infrared illumination, audio, or other sensors in a hybrid product still require separate disclosure and assessment. Mounts, enclosures, power supplies, adapters, and cables should prevent falling, overheating, entanglement, and obstruction of bed exit or care. Institutional settings also require suitable cleaning and mounting load. Electrical, network, or mechanical safety conformity does not establish temperature accuracy or medical performance of respiration and sleep algorithms.
Maintenance, medical limits, and response
Routine checks should cover cleanliness of the lens and protective window, focus, position, empty-bed thermal background, clock, storage, network, and the alert path. Calibration, defective-pixel handling, and software updates should follow manufacturer instructions. Seasonal conditions, climate control, bedding, heating equipment, occupants, firmware, and algorithm changes should be recorded and followed by retesting. A whole-image temperature jump, palette shift, fixed hot spot, repeated frame, implausibly smooth trace, or conflict with observation should first be investigated as an imaging or data-chain fault.
A general-wellness system may support review of movement, posture, respiratory trends, and surface-temperature change, but it cannot replace polysomnography, a validated home sleep-apnoea test, contact vital-sign monitoring, or clinical temperature measurement. A claim to diagnose, screen, continuously monitor, or trigger treatment requires applicable medical-device authorization in the market of sale, an intended population, performance evidence, contraindications, and human review. “Passive,” “contactless,” and “no light required” do not confer medical-grade performance.
A normal display cannot exclude real symptoms, while an unusual thermogram should not lead to self-treatment without clinical context. Witnessed breathing pauses, awakening while choking, marked daytime sleepiness, persistent fever, chest pain, fainting, difficulty breathing, or another acute concern should lead to medical or emergency help based on the symptoms. An infant-facing thermal product cannot replace adult supervision or safe-sleep practice. Without evaluation and authorization for the relevant purpose, it should not claim reliable vital-sign monitoring or prevention of sudden infant death.