Technical Overview v1.0

Sensor-based early detection and outbreak surveillance for cattle · replaces the February 2026 whitepaper in full

BovRes BR-1: Technical Overview

Sensor-based early detection and outbreak surveillance for cattle | v1.0, September 2026 | For veterinary, engineering and due-diligence readers

This document replaces the February 2026 "Technical Whitepaper" in full. It describes what the system measures, why those measurements are expected to detect disease early, how the system is built, and how its performance will be validated. It contains no performance claims; those will follow from the studies in Section 6.

1. Detection rationale

1.1 Foot-and-mouth disease timeline in cattle

Controlled challenge studies published in 2025 give the clearest picture to date of the pre-clinical phase in cattle. Following exposure, FMDV RNA was detectable in nasal secretions from 24 hours in some animals and from 72 hours in all animals that went on to develop clinical disease; viraemia was detected from 72 hours; the first clinical sign, either fever above 40 °C or an oral vesicle, appeared at approximately 96 hours; and infected cattle transmitted to naive in-contact cattle at least 24 hours before their own clinical signs. Experimental incubation ranges from 2 to 7 days and is more variable in the field. Fever is consistently the earliest clinical sign and precedes lesion formation. Earlier infrared thermography work confirmed fever as an effective screening signal for FMD, with eye temperature a useful proxy for core temperature.

1.2 What sensors can see before a person can

Temperature. Reticulorumen temperature boluses detect hyperthermia 12–136 hours before clinical signs of bovine respiratory disease (BRD) in young bulls. Continuous surface or tympanic temperature with ambient compensation and per-animal circadian baselines reproduces much of this signal non-invasively.
Behaviour. Accelerometer-derived feeding, rumination, lying and activity measures change up to 5–7 days before BRD is clinically apparent in calves and are the basis of commercially validated products (SenseHub Feedlot is documented to identify at-risk cattle before clinical signs). FMD produces lameness, reluctance to stand, salivation and reduced intake, all of which have accelerometer signatures.
Herd context. A single febrile animal is common; several animals on one farm developing fever and activity collapse within 24–48 hours of one another is not. Detecting the cluster, not just the individual, is the surveillance signal that matters for FMD and is not offered by per-cow productivity systems.

1.3 Design hypothesis

Fusion of continuous temperature and behaviour signals against per-animal learned baselines, combined with herd-level anomaly clustering, will flag FMD, BRD, mastitis and lameness cases at least 24 hours (target: 48 hours) before routine visual detection, with sensitivity ≥80% and specificity ≥90% for clinical illness. This hypothesis will be tested, not assumed.

2. System architecture

LayerComponentDescriptionSensingBR-1 collarContact IR thermopile (MLX90614 class) at the neck with ambient reference; 3-axis accelerometer at 25 Hz; optional GPS; optional tympanic ear-tag sensor for higher-fidelity core-temperature proxyEdge processingESP32-class MCUFeature extraction (temperature trend, activity index, lying bouts, rumination proxy from jaw/neck motion), per-animal rolling baseline, anomaly score; compact 15-minute uplink; raw windows on requestPowerFlexible solar + kinetic harvester + LiFePO4/supercapacitor bufferDesign target: no battery service over a 5-year life; low-power sleep between samplesConnectivityLoRaWAN 868 MHz (ICASA licence-exempt)Collar-to-collar relay in a mesh, extending range across large paddocks; solar gateway with cellular or satellite backhaul and store-and-forwardCloudSouth African-hosted time-series platformPostgreSQL/TimescaleDB, Python analytics service, n8n for workflow automation and alert routing, REST API, POPIA-compliantApplicationsHerdLink Farm, Surveillance, AssuranceMobile PWA for farms; district dashboards for state vets; verification API for processors and insurersInteroperabilityLITS-SA, DALRRDAnimal identifiers and movement records in LITS-SA format; surveillance exports structured to WOAH Chapter 8.8 clinical-surveillance evidence

3. Hardware

Housing: IP68 ASA/TPU, UV-stabilised, radiused for rub resistance; strap with anti-rotation profile keeps the sensor pod in contact at the ventral neck; field-replaceable strap.
Antenna: graphene-enhanced conductive ribbon (SabiNano) integrated in the strap for low-profile 868 MHz performance; conventional copper alternative qualified in parallel to de-risk supply.
Bill of materials target: R1,250 at 100 units, R650 at 45,000 units; injection moulding, PCB assembly and final assembly in Cape Town/Gauteng.
Bench qualification (Stage 0): temperature accuracy vs reference thermometry across 0–45 °C ambient; immersion, UV and abrasion; 60-day power-budget test in winter light; mesh range and packet-loss tests on an extensive farm.

4. Analytics

4.1 Per-animal baselines

Each animal's temperature and activity follow strong circadian and management-driven cycles. The collar maintains a rolling 7–14 day baseline per hour-of-day and reports deviations in standardised units, which makes alerts robust to breed, season, weather and milking schedule without hand-tuned thresholds. Night-time windows (low ambient temperature, low activity) are weighted more heavily for fever detection because surface-temperature measurements are most stable then.

4.2 Multi-signal case definitions

Alerts are raised from combinations of signals mapped to veterinary case definitions, for example: sustained temperature deviation plus activity decline plus rumination decline (systemic illness); asymmetric activity and lying pattern (lameness); temperature deviation with normal activity (early febrile phase, priority check). Case definitions are versioned and reviewed with the veterinary partner.

4.3 Herd- and district-level inference

Mesh proximity data yields a contact network. Clusters of individual anomalies that are close in time and in the contact network raise a herd alert; herd alerts across neighbouring farms raise a district alert to the provincial veterinary service. This is the surveillance product and is the primary FMD-specific innovation.

4.4 Machine learning roadmap

The initial system uses interpretable statistical baselines and rules, which are auditable by veterinarians and require no training data. Labelled data from the validation study and pilots (clinical scores, treatments, laboratory results) will be used to train supervised models for disease-specific classification and lead-time optimisation, with performance reported in a peer-reviewed publication.

5. Regulatory and ethical

Radio: ICASA type approval for the LoRa module and final device; SABS/IEC EMC and electrical safety testing.
No therapeutic claim: the device is not a stock remedy or veterinary medicine and does not fall under Act 36 of 1947 or the Medicines Act.
Animal welfare: collar mass, fit and behaviour impact reviewed by the partner faculty's animal ethics committee before field deployment.
Data: POPIA; farm data owned by the farmer; aggregated surveillance data shared with DALRRD under agreement; hosted in South Africa.

6. Validation programme

StageTimingDesignPrimary outcomes0. BenchQ4 202630 collars; lab and one extensive farmSensor accuracy; durability; power; mesh range1. Field baselineQ1 2027100 animals (50 dairy, 50 feedlot), 12 weeks; daily blinded clinical scoring by faculty staff; treatment records as referenceSensitivity, specificity, lead time vs visual detection for BRD, mastitis, lameness, any illness2. FMD evaluationQ2–Q3 2027Farms under FMD restriction in KZN/EC with provincial vets; confirmed cases (clinical + laboratory) as referenceProportion of confirmed cases flagged before clinical reporting; lead time; false-alert rate3. Surveillance pilotQ4 20272,000+ animals in one district; dashboard in use by state vetsUsability; time-to-action; contribution to surveillance documentation

Protocols will be pre-registered, ethics-approved and run by the faculty partner as principal investigator. Results will be published regardless of outcome.

7. Intellectual property

A South African provisional patent application has been filed (Global IP Co, ref. PT/PV/ZA00002396). The complete specification, due within 12 months of the provisional date, will be drafted around: (1) the self-powered, mesh-relaying collar architecture; (2) the on-device per-animal baseline and anomaly method; (3) the contact-network-based herd and district outbreak inference method; (4) the strap-integrated conductive antenna. Claims relating to therapeutic emission in the provisional filing will not be carried forward.

8. Key references

Arzt J. et al. Incubation phase transmission of foot-and-mouth disease virus in cattle: experimental evidence and simulated impacts. Sci Rep (2025). https://www.nature.com/articles/s41598-025-34132-x

Epidemiologic consequences of preclinical transmission of FMDV in cattle. Front Vet Sci (2025). https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2025.1651091/full

The pathogenesis of foot-and-mouth disease virus: current understandings and knowledge gaps. Vet Res (2025). https://veterinaryresearch.biomedcentral.com/articles/10.1186/s13567-025-01545-5

Rainwater-Lovett K. et al. Detection of FMDV infected cattle using infrared thermography. Vet J (2009). Normal variation in thermal radiated temperature in cattle: implications for FMD detection. BMC Vet Res (2011). https://pmc.ncbi.nlm.nih.gov/articles/PMC3235061/

Timsit E. et al. Early detection of bovine respiratory disease in young bulls using reticulo-rumen temperature boluses. Vet J (2011). https://pubmed.ncbi.nlm.nih.gov/20947394/

Detection of physiological changes using a triaxial accelerometer and temperature sensor-equipped bolus biosensor in calves. Animals (2024). https://doi.org/10.3390/ani14192815

Wearable collar technologies for dairy cows: a systematized review of current applications and future innovations in PLF. Animals 15(3):458 (2025). https://www.mdpi.com/2076-2615/15/3/458

Design of an intelligent wearable device for real-time cattle health monitoring. Front Robot AI (2024). https://www.frontiersin.org/journals/robotics-and-ai/articles/10.3389/frobt.2024.1441960/full

Merck Animal Health. SenseHub Feedlot. https://www.merck-animal-health-usa.com/species/cattle/sensehub

WOAH Terrestrial Animal Health Code, Chapter 8.8 Infection with foot and mouth disease virus. https://www.woah.org/fileadmin/Home/eng/Health_standards/tahc/2024/en_chapitre_fmd.htm

Confidential. Provided under NDA to prospective investors and partners of Burmain Holdings (Pty) Ltd. Not for onward distribution. This document makes no therapeutic claims; BR-1 is a monitoring and decision-support device. Performance figures will be reported from independent, pre-registered studies.