Retraining of traving dogs for the detection of COVID-19 human patients

12 November 2020 → 28 February 2021
Federal funding: various
Research disciplines
  • Medical and health sciences
    • Health management
    • Preventive medicine
    • Health and community services
Project description

During the first outbreak of COVID-19 in Europe begin 2020, Prof. Dr. Dominique Grandjean from the faculty of veterinary medicine Maison Alfort but also colonel in the fire brigade, Paris, took the initiative to investigate if tracing dogs could made the difference between CPVID 19 – and + patients using sweat samples obtained from the inguinal and armpits regions (30 minutes of sampling). Since the virus is not excreted in sweat, this was accepted as a safe method for dogs and humans. The preliminary results were published begin December 2020. Meanwhile, other countries had similar set ups. Results were reported with an accuracy of 90 %, sensitivity of 100 % and success rate between 83 and 100 %. Up to now, 5 papers are published in international journals. Operational dogs were already used in airports/border controls in Finland, Chili, UAE and Lebanon whereby sweat samples from traveling persons were presented to the dogs.

The aim of the present project was to investigated of tracing dogs from Belgium could also made the difference between COVID19 – and + patients.


A steering group was founded with the following persons:

  • Dr. F. Gasthuys, Faculty of Veterinary Medicine, UGent
  • Dr. H. Guyot, Faculty of Veterinary Medicine, ULiège
  • Ir. Chris Callewaert, Faculteit Bio-engeneer Sciences, UGent
  • Miguel Stevens, past president International Working Dog Breeding Association.
  • Fits chief officer Rony Vandaele, director Dog Support Federal Police


In order to stream all people involved in tracing dogs (Defense, Fire Brigade, Federal Police, Civil Protection) a meeting was set up in May 2020 to join forces. A protocol including an estimation of the budget was formulated whereby all trainings were centralized in the “Centrum Hondensteun” of the Federal Police in Neerhespen. Additionally, it was decided to work in 2 phases:

  • Phase 1 to validate a method and protocol with a budget of 60.000 (material, analyses etc.)
  • Phase 2 to train suitable dogs for operational purposes.


The Federal Government approved the project half November 2020 whereby a Royal Arrest had to be made to provide the budget. UGent opened a credit line in order to allow the start up of the project until the Arrest was published. The Arrest was finally published end of January.


Sweat samples from COVID19 + patients were obtained from more than 20 human hospitals, elderly institutes and RVT’s. Negative samples were also collected in 550 volunteers (confirmed by PCR) from the service club KIWANIS. All samples were stored and transported at deep freeze temperature.


All dogs included in phase 1 of the project had already a pretraining as tracing dogs (explosives, drugs, search and rescue). Half December a group of 15 dogs with their personal trainer was available from the different official bodies (Federal Police, Fire Brigade, Defense, Civil Protection).

A first selection was done whereby 11 dogs and trainers were able to continue the training. Several problems were encountered whereby the protocol needed to be adjusted. The multiple use of one sample was confusing for the dogs: the decision was made to present one sample maximally 2 times. Consequently, the number of sample needed to train the dogs increase significantly (600 + samples for 11 dogs per week). Mid-January, the number of dogs was further reduced to allow multiple training runs per day similar tot the training in France.

From Mid-December to Mid-January only + samples were presented to the dogs to imprint the odor. In the first half of January both – and + samples were used in the training. The dogs were able to distinguish clearly the different samples, supporting the thesis of a specific Volatile Organic Compounds present in the sweat samples of COVID19 patients.

End of January the number of dogs was further reduced to 6 to allow an intensification of the training without blanc samples and only with + and – sweat samples. Additionally, the dogs tracing explosives were superior compared to the search and rescue digs.

Beginning of February the dogs were working in teams of 2 whereby one sample was presented to both dogs of the same team. After 4 weeks on an intensive training, all dogs were able to make the difference between the samples; however, further trainings session are required to optimize the precision of the dogs. It became also clear that an ad rem sampling method of the sweat samples was crucial to obtain acceptable training results of the dogs. Even more, the use of a standard carrier for the sweat samples became clear. The steering group decided to use only one single kind of cotton balls sealed in uniform plastic bags (no gauze nor silicone carriers). The initial package of glass jars was abandoned because of pure practical reasons (storage and transport of the volume of jars

A preliminary validation took place in 6 dogs on 2 different days:

  • 01/02 Sensitivitity: 70%                   Specificity: 80%
  • 02/20 Sensitivitity: 80 -100%          Specificity: 90-100%

The final validation is planned in the week of 22/02 whereby all but one collaboarators  are blinded fort he samples. The availability of positive samples is not always assured because of several reasons (high working load in IC units, refusal of less sick people for donation of samples etc.). Consequently, the intensive training of the dogs can be suboptimal. It is essential to determine the specific volatile organic compounds in the sweat which can be done only by highly specialized lab’s including the one of the UV San Diego. The steering group has close contacts with this lab and an budget was reserved in order to look for these VOC’S.