The gap we were built to close
Dogs are stoic by nature. Hardwired over millennia to suppress outward expressions of pain as a survival mechanism, companion dogs routinely mask discomfort long past the point where a condition has meaningfully progressed. Owners miss it. Veterinarians, seeing a patient for minutes at a time, weeks or months apart, miss it too. By the time pain becomes visible, the window for early, less invasive intervention has often already closed.
Tether has built a system designed to operate inside that window. The Ruff is a collar-mounted wearable device that continuously captures five streams of physiological and behavioral data from a dog during its ordinary daily life. That data flows to the Tether cloud platform, where an Individualized Longitudinal Baseline Engine (ILBE) compares real-time sensor outputs against a behavioral model built from that specific animal's own history, not population averages, not breed norms. When meaningful deviation is detected, the system generates a Pain Probability Index (PPI), classifies the probable character and anatomical locus of the pain, ranks candidate conditions, and delivers urgency-stratified guidance to both owners and veterinarians.
This document describes what the system measures, how it works, what it produces, and, critically, what it does not claim to do. The system does not diagnose. It does not replace physical examination, imaging, or clinical judgment. It identifies the probability of a pain state and supports the humans responsible for acting on that information.
The goal is not to replace veterinary care. The goal is to ensure that by the time a dog reaches the clinic, the veterinarian has weeks of objective, continuous behavioral data, not a five-minute observation window.
The invisibility of canine pain
Pain concealment in dogs is not a behavioral quirk. It is an evolutionary adaptation rooted in predator avoidance: an animal that visibly limps or vocalizes distress advertises vulnerability. The instinct to suppress outward pain expression is deeply hardwired, and it operates regardless of the severity of the underlying condition. A dog with moderate osteoarthritis may move normally enough during a clinic visit, adrenaline, changed environment, and the instinct to appear capable all contribute, while spending the other 23 hours of its day struggling to rise from rest, hesitating before stairs, waking repeatedly through the night.
The validated clinical pain scales used in veterinary practice, the Canine Brief Pain Inventory (CBPI), the Glasgow Composite Measure Pain Scale Short Form (CMPS-SF), the Helsinki Chronic Pain Index (HCPI), are well-constructed instruments. They are also episodic, observer-dependent, and administered at intervals of weeks to months. They were designed for the clinic visit, which means they are structurally blind to the vast majority of a dog's lived experience with pain.
Veterinary costs have increased by 55.5% since 2019, while clinic visit volume has declined for four consecutive years. Owners are paying more per visit and attending fewer of them. The result is a growing population of dogs whose pain conditions are either undetected between visits or detected late, at the point of emergency presentation, when intervention is more intensive, more expensive, and less likely to fully restore function. Early detection moves diagnosis upstream toward a point where conservative management, targeted physiotherapy, or timely pharmaceutical intervention can meaningfully alter the course of the condition.
The four gaps existing products don't close
Consumer pet wearables have advanced meaningfully over the past decade. GPS tracking, step counting, and basic activity monitoring are now widely available. A smaller number of products have extended into vital sign monitoring. These systems represent real progress. They also share four structural limitations that prevent them from serving as reliable early pain detection tools.
POPULATION THRESHOLDS CANNOT ACCOUNT FOR INDIVIDUAL VARIATION
Dogs vary enormously in their baseline physiological signatures. A healthy adult Greyhound has a naturally lower resting heart rate than a healthy adult Chihuahua. A brachycephalic breed has respiratory patterns that would be flagged as abnormal in a mesocephalic dog. Systems that compare sensor outputs against population averages generate false positives in some animals and miss genuine signals in others. The appropriate comparison is not against the breed or population, it is against that individual dog's own established normal.
Pain-associated vocalizations, whimpering, groaning, yelping, are among the most specific behavioral indicators of pain in dogs. Yet prior wearable systems have not treated acoustic data as a primary pain biomarker stream. This leaves a significant dimension of pain-related behavioral expression unmeasured.
Fine muscle micro-tremors, the involuntary neuromuscular tension associated with pain-induced muscle guarding, are not detectable by motion sensors or temperature sensors. They require a sensor capable of detecting sub-millimeter displacement: millimeter-wave radar. This capability has not, prior to this platform, been incorporated into a collar-mounted wearable for continuous pain monitoring.
A pain score, on its own, is not enough. An owner who receives a score of 71 out of 100 needs to know what kind of pain is likely present, what is probably causing it, how urgent the situation is, and what to do next. Without a contextualization layer, a detection system produces anxiety without guidance.
Five complementary sensors
The Ruff is a collar-mounted device housing five sensor components selected for their complementary, non-redundant coverage of the physiological and behavioral dimensions most relevant to canine pain. The device generates in excess of 10,000 data points per animal per day, transmitting continuously to the Tether cloud platform via encrypted wireless protocols. It is waterproofed to IPX7 standard and designed for continuous wear. GPS-based location tracking is not included in the initial hardware release and is planned for inclusion in future device updates as part of Tether's ongoing hardware investment.
| Sensor / Component | What It Captures | Clinical Relevance |
|---|---|---|
| 60 GHz Radar (Acconeer A121) | Respiration rate, heart rate, micro-tremor index, contactless | Primary non-contact physiological channel. Works through the enclosure and through fur. Micro-tremor index detects muscle guarding, neuromuscular tension, and pain-related tremor, a capability with no equivalent in prior wearable systems. Values interpreted as trends against the dog's own baseline. |
| 6-Axis IMU, Primary (LSM6DSV16X, U1) | Gait symmetry, head-bob, activity level, postural transitions, restlessness | Always-on sentinel in the firmware architecture. Drives activity level, gait symmetry, posture transitions. Gait asymmetry and head-bob are primary lameness indicators. Slow postural transitions are a hallmark of arthritis and spinal pain. |
| Dedicated Vibration Channel (LSM6DSV16X, U10) | High-rate vibration on an isolated signal path | A second IMU on its own SPI interface, providing fine vibration-domain features independent of the primary motion stream. Captures physiological information that standard motion sampling at lower rates cannot. |
| MEMS Microphones ×2 (SPH0641LU4H-1, MK1 & MK2) | Vocalizations, panting, coughing, classified on-device. Raw audio never transmitted. | Paired placement supports robust capture on a moving animal. On-device firmware converts digital streams into audio features. Vocalization contextualization that helps distinguish fear, anxiety, and pain is available from launch. |
| Temperature (TMP117, U2) | Surface temperature via dedicated thermal path through enclosure | Accuracy in the 0.1°C class, sufficient to distinguish genuine physiological shifts from sensor noise. Temperature shifts accompany inflammation, infection, and stress responses. Signal value comes from trend deviation against the individual animal's own baseline. |
| GPS, planned | Not in v1 hardware | Roaming radius, daily distance, planned for future hardware updates. These GPS-derived metrics will contribute to the behavioral sub-score when introduced. |
Every dog is its own control
The Individualized Longitudinal Baseline Engine is the core technical innovation of the Tether platform. Rather than comparing a dog's sensor outputs against breed averages or population reference ranges, the ILBE constructs and maintains a continuously updated behavioral model built from that specific dog's own historical data. Every dog receives a baseline that is uniquely theirs. Every deviation is measured against the version of themselves the system has learned over weeks and months of continuous wear.
Pain in dogs does not present as a single threshold violation. It presents as correlated drift across multiple subtle signals over time, gait symmetry shifts by a few percent, resting respiratory rate creeps upward, vocalization frequency moves into pain-associated bands, morning activity onset slows. No individual signal is diagnostic. The multi-signal deviation pattern, measured against an individualized baseline, is.
, The 24-hour baseline detects acute changes, sudden activity drops, gait symmetry collapses, fever spikes, and vocalization bursts consistent with acute injury or acute illness.
, The 7-day baseline identifies sub-acute trends developing over days, such as a progressing soft-tissue injury or the early stages of an infectious process.
, The 30-day baseline represents the dog's established normal and is the primary reference for chronic pain detection, the gradual, insidious drift that characterizes osteoarthritis, degenerative joint disease, and progressive neurological conditions.
Baselines require approximately 14 days before they are considered reliable. Confidence scores, present on every assessment, reflect baseline maturity explicitly. Breed, age, and morphological context are incorporated into what the ILBE considers normal for each individual, preventing false positives from breed-normal physiology while remaining sensitive to genuine deviations.
From signals to clinical output
The ILBE's output feeds a scoring engine that computes a composite pain score on a 0, 100 scale, derived from three weighted sub-scores:
, Behavioral score (40% weight)
Nighttime restlessness, activity trend slope, vocalization rate and intensity, and activity reduction. Behavioral changes are the broadest indicator across pain types and receive the highest weight.
, Mobility score (35% weight)
Activity level, gait symmetry, peak angular velocity, head-bob asymmetry (forelimb lameness indicator), postural transition rate and duration, and morning stiffness index.
, Cardiorespiratory score (25% weight)
Heart rate elevation (radar-derived), respiratory rate elevation (radar-derived), micro-tremor index, circadian HR pattern disruption, resting panting rate, coughing rate, and skin temperature deviation.
Each sub-score is driven by the single worst indicator within its category, not an average. The composite score is classified into a Pain Probability Index (PPI) tier: Green (no significant indicators), Amber (monitoring or veterinary consultation recommended), and Red (veterinary attention recommended).
Not just a number, what it means and what to do
Each assessment classifies the detected pain state across four clinical dimensions simultaneously. Pain acuity describes the temporal character, acute, subacute, chronic, or episodic. Anatomical locus uses kinematic signature patterns to estimate the probable body region involved. Pain character classifies the probable pathophysiological mechanism, nociceptive, inflammatory, neuropathic, visceral, or mixed. Environmental and activity context is integrated explicitly to prevent false-positive classifications.
The causal attribution model produces a ranked differential list of probable conditions weighted by breed-specific condition prevalence, age-adjusted risk, sensor signal pattern matching, and the individual animal's known health history. This is not a clinical conclusion. It is a data-backed starting hypothesis intended to direct clinical attention toward the most likely diagnoses before the examination begins.
The recommendation engine maps the pain classification and differential output to an urgency tier, each with a distinct recommended response protocol. Recommendations are voiced separately for owners in plain language and for veterinarians in clinical language, including examination focus areas, imaging modalities to consider, and treatment response benchmarks.
Every assessment generates a SOAP-formatted veterinary report structured to align with the format used in veterinary medical records. This transforms the clinic visit from a single-point snapshot into a longitudinal consultation. The veterinarian arrives having seen not just today's presentation but the behavioral arc of the preceding weeks.
What this system does not do
Clinical credibility requires clarity about limitations. These are accurate characterizations of a system in active development, described here so that veterinary professionals, owners, and all others evaluating this platform can form an accurate understanding of what it can and cannot tell them.
, It does not diagnose. Differential rankings are probabilistic suggestions to guide examination, not clinical conclusions. Veterinary examination is always recommended for elevated scores.
, It does not replace physical examination. Palpation, manipulation, response to touch, gait assessment, and imaging provide information no wearable sensor can supply.
, It requires a maturation period. Reliable baselines require approximately 14 days of continuous wear. First-week results carry reduced confidence, clearly indicated in the output.
, It is strongest for musculoskeletal and behavioral pain. Dental and oral pain, and some visceral pain presentations, may have subtler sensor signatures.
, Audio classification has known limitations. On-device vocalization classification may occasionally confuse environmental sounds with pain-associated vocalizations. Multi-dog households present attribution challenges.
, Skin temperature is surface temperature, not core body temperature. The TMP117 measures neck surface temperature through the collar's thermal path. Clinical value comes from trend deviation against the individual animal's own baseline.
, Clinical scale mappings are estimated, not validated. CBPI, CMPS-SF and other clinical metrology instrument equivalents are sensor-derived estimates requiring prospective calibration studies before carrying clinical authority.
, GPS-derived mobility metrics are not available in the initial hardware release. Roaming radius and daily distance will be introduced in future hardware updates.
The path to clinical credibility
The system described in this document is a clinically reasoned design grounded in established veterinary pain biology. It is not yet a clinically validated system. Specific open questions requiring prospective clinical data include: scoring weight validation across pain types; threshold calibration for mild, significant, and severe deviation categories; brachycephalic respiratory rate multiplier calibration; and CCD versus nocturnal pain discrimination in geriatric dogs.
Tether is actively seeking veterinary partners, individual practitioners, academic clinical programs, and specialist referral centers, to participate in the prospective validation studies needed to answer these questions. If you are a veterinary professional with relevant clinical experience, we invite your engagement.
| Sensor | Hardware | Sample Rate | Derived Outputs |
|---|---|---|---|
| Radar, physiological | Acconeer A121 (60 GHz) | Every 10s | Respiration rate, heart rate, micro-tremor index |
| Movement, primary | LSM6DSV16X IMU (U1) | Every 1s | Gait symmetry, activity level, head-bob, morning stiffness, restlessness, postural transitions |
| Movement, vibration channel | LSM6DSV16X (U10), isolated SPI | High-rate | Fine vibration signatures, physiological features in development |
| Vocalization (×2 microphones) | SPH0641LU4H-1 (MK1 & MK2) | Every 2s | Vocalization rate, panting rate, coughing rate, vocal intensity |
| Temperature | TMP117 (U2) | Every 2s | Surface temperature deviation from individual baseline |
| GPS, planned | Not in v1 hardware | Roaming radius, daily distance, planned for future updates |
| Parameter | Value | Purpose |
|---|---|---|
| Behavioral sub-score weight | 0.40 | Composite score |
| Mobility sub-score weight | 0.35 | Composite score |
| Cardiorespiratory sub-score weight | 0.25 | Composite score |
| Acute floor, 1 flag | 55 | Acute pathway |
| Acute floor, 2+ flags | 75 | Acute pathway |
| EWMA alpha: short-term (24h) | 0.30 | Baseline update |
| EWMA alpha: medium-term (7d) | 0.10 | Baseline update |
| EWMA alpha: long-term (30d) | 0.03 | Baseline update |
| PPI Green threshold | < 0.30 | PPI tier |
| PPI Amber threshold | 0.30, 0.65 | PPI tier |
| PPI Red threshold | ≥ 0.66 | PPI tier |
| Baseline maturity (full confidence) | 14 days | Confidence scoring |
| Minimum confidence floor | 0.10 | Confidence clamp |
CBPI (Canine Brief Pain Inventory): A validated owner-reported clinical pain scale for dogs, producing a Pain Severity Score and a Pain Interference Score on a 0, 10 scale.
CMPS-SF (Glasgow Composite Measure Pain Scale Short Form): A validated clinician-administered acute pain scale for dogs with an intervention threshold at a score of 6 or above.
EWMA (Exponentially Weighted Moving Average): A statistical method for updating a running average that gives more weight to recent observations, used to maintain rolling baselines for each sensor metric.
HRV (Heart Rate Variability): The variation in time between successive heartbeats, measured as RMSSD. Lower HRV is associated with pain, stress, and autonomic nervous system activation.
ILBE (Individualized Longitudinal Baseline Engine): Tether's core detection system, which builds and maintains a behavioral baseline model specific to each individual dog.
IVDD (Intervertebral Disc Disease): A common canine spinal condition involving disc herniation, particularly prevalent in chondrodystrophic breeds.
Micro-tremor index: A measure of fine involuntary muscle oscillations detected by the 60 GHz radar, associated with pain-induced muscle guarding and neuromuscular conditions.
PPI (Pain Probability Index): The composite output of the ILBE scoring engine, expressed as a value from 0 to 1 and classified into three tiers: Green, Amber, and Red.
SOAP (Subjective, Objective, Assessment, Plan): A standard format for clinical documentation in veterinary and human medicine.
TMP117: Texas Instruments digital temperature sensor (0.1°C accuracy class) used in The Ruff to measure surface temperature through a gold-plated thermal path in the collar enclosure.
1. International Association for the Study of Pain Task Force on Taxonomy. In: Merskey H, Bogduk N, eds. Classification of Pain. 2nd ed. Seattle: IASP Press, 1994:209, 214.
2. Merck Veterinary Manual. Recognition and Assessment of Pain in Animals. Reviewed/Revised March 2023.
3. Walsh S. Understanding the link between canine pain and problem behaviours. Veterinary Ireland Journal. February 2025.
4. Mills DS, Demontigny-Bedard I, Gruen M, et al. Pain and Problem Behavior in Cats and Dogs. Animals (Basel). 2020;10(2):318.
5. Barcelos A-M, Mills DS, Zulch H. Clinical indicators of occult musculoskeletal pain in aggressive dogs. Veterinary Record. 2015;176(18):465.
6. Mills D, Zulch H. Veterinary assessment of behavior cases in cats and dogs. In Practice. 2023;45:444-458.
7. Landsberg G, Hunthausen W, Ackerman L. Behavior problems of the dog and cat. Vet Clin Small Anim Pract. 2013;43(5):975-1003.
8. Cornell Richard P. Riney Canine Health Center. Cohen A. Recognizing Pain in Dogs. vet.cornell.edu. Updated October 2025.
9. Mota-Rojas D et al. Current Advances in Assessment of Dog's Emotions, Facial Expressions, and Their Use for Clinical Recognition of Pain. Animals. 2021;11(11):3334.
10. Demirtas A et al. Dog owners' recognition of pain-related behavioral changes in their dogs. Journal of Veterinary Behavior. 2023;62:12-17.
11. American Animal Hospital Association. AAHA 2022 Pain Management Guidelines for Dogs and Cats. aaha.org.
12. Caddiell RM et al. Pain sensitivity differs between dog breeds but not in the way veterinarians believe. Frontiers in Pain Research. 2023;4:1165340.
13. Brown DC et al. The use of the Canine Brief Pain Inventory to characterize osteoarthritis-associated chronic pain in dogs. J Vet Intern Med. 2007;21(2):248-255.
14. Reid J et al. Development of the short-form Glasgow Composite Measure Pain Scale (CMPS-SF). Vet Anaesth Analg. 2007;34(2):97-104.
15. Downing R, Della Rocca G. Pain in Pets: Beyond Physiology. Animals (Basel). 2023;13(3):355.
16. Gibbons J. Petflation 2026, March Update: All Pet Prices Reach Record Highs. Pet Business Professor. April 20, 2026.