Automatic disease detection and monitoring in calves

Lead Research Organisation: University of Bristol
Department Name: Experimental Psychology

Abstract

Bovine respiratory disease (BRD) is the most common and costly disease affecting cattle in the world. BRD is a complex bacterial infection that can be fatal and is estimated to cost the UK cattle industry £80M annually. Although manual scoring systems exist to aid early identification of the disease, they are time consuming and rarely used in practice. Commonly, identification of BRD is only in later stages of the disease when antibiotics are essential for treatment. Early and automated identification of BRD will have significant impact: 1) on the economic cost to farmers; 2) reducing the quantity of antimicrobial medicines used to treat the disease; and 3) improving the general welfare of animals.

The proposed project uses artificial intelligence techniques, coupled with visible-range and thermal cameras, to identify BRD at the earliest possible stage with main goals of establishing: 1) how early in disease development affected animals can be reliably identified; 2) the best way to scale up image capture and machine learning to automatically screen animals and alert farmers to those needing treatment; together with 3) developing a protocol for effective use of the trained system. The aim of the proposal is to develop a system based on providing the best possible information in a timely manner, which is key to making right judgements for farmers and vets alike. It is believed that a system based on low cost cameras and sensors, together with state of the art deep neural networks, can provide this.

The completed fellowship will result in a working and tested prototype system capable of development into a viable commercial product. During the fellowship a network of industry collaborators (including farmers, vets, advisory/regulatory bodies, equipment manufacturers and food producers) will be developed to support and promote the research and resulting product.

Planned Impact

This proposal details of the activities needed for early automatic detection of disease in collaboration with the farming industry. In the first instance and for the purposes of research and development, the focus of the proposal is Bovine respiratory disease (BRD). BRD is a complex bacterial infection and is the most common and costly disease affecting cattle. In particular, planned collaboration with UK industry in the form of commercial farms and vets, will be developed in order to support the programme outlined in the proposal. Completion of the programme not only offers opportunities within the UK, based on a commercial start-up, but also the potential to develop the product for export, both of which will be important for the UK's future industrial strategy. Beneficiaries of the proposed programme are in both the academic and professional communities, as well benefits for society more generally.

Academically, benefits will accrue in artificial intellegence through applying deep learning to a combination of (low quality) visual and thermographic spectrum data, with the aim of characterisation of pre-clinical BRD. Of particular interest will be the potential to identify which stage an animal is in the disease development, rather than a simple binary classification. Indeed, it is expected that this will highlight subtleties and relationships in the data that are not currently known and will aid disease characterisation. The ability to characterise disease in this 'automatic' way will also be of considerable interest to academic vets who study animal diseases. In veterinary medicine one of the deliverables will be a method of screen-based scoring of disease, based on assessment of visible-range and thermographic images, which will further contribute to the general approach to study diseases in animals. Overall and significantly, the project will contribute to knowledge of how early in the development of disease (BRD in the first instance) it can be reliably identified, this in turn provides information for vets when choosing the most appropriate treatment.

Being able to identify disease earlier has obvious benefits to farmers, such as fewer sick animals for shorter periods, that contribute to lower costs and better welfare. Professional veterinary surgeons also benefit from an addition to the diagnostic tools at their disposal, together with more appropriate treatment. Of particular interest and benefit from the point of view of professional vets is that early diagnosis leads to reduced use of antimicrobial medicines. This is a significant and pressing issue that leads to benefits to wider society through reducing the levels of antimicrobial resistance.

These benefits have already been acknowledged by our collaborators as the following quote (from an email to the applicants, talking about BRD) illustrates

"In the last few weeks - at least in my practice - initial treatment was often followed by re-occurrence and in such cases the previous treatment was not effective anymore. This leads to a prolonged treatment period, which increases the amount of medicine required.

A repeated and prolonged treatment profile regularly results in side-effects; particularly problems with digestion, anorexia, increased gas levels in the rumen, diarrhoea, changes in gastric flora, enterotoxemia, pododermatitis, lameness, and recumbency followed by emaciation. These individuals generally fall behind, others push them away from feeders and harass them, which leads to continuous fear and stress, loss of growth, and increased mortality due to other diseases.

These are real issues that I experience on a daily basis, causing significant amount of loss for the farmers, not to mention aspects of animal welfare. The proposed early detection coupled with timely treatment promises lower medicine consumption, which will minimise the side-effects of antimicrobial treatment."

Publications

10 25 50
 
Description The sensory platform, designed and built as the first objective the project, has been achieved and implemented on site (two large commercial dairy farms in the southwest of England). Multiple sensory units are operating automatically and continuously, delivering images to our computers in Bristol, where they are processed.
Exploitation Route A stable and reliable data-flow is essential to achieving the remaining objectives for the project.
Sectors Agriculture, Food and Drink

 
Description GPU Grant Program
Amount $1,500 (USD)
Organisation NVIDIA 
Sector Private
Country Global
Start 11/2018 
End 05/2021
 
Description Impact Acceleration Account: Automatic disease detection in racehorses using low-cost thermography and machine learning
Amount £4,559 (GBP)
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Public
Country United Kingdom
Start 07/2019 
End 12/2019
 
Description School of Psychological Science Research Committee Undergraduate Vacation Studentship
Amount £1,200 (GBP)
Organisation University of Bristol 
Sector Academic/University
Country United Kingdom
Start 06/2019 
End 09/2019
 
Description Advisory Board meeting 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact Organised and held initial Advisory Group meeting with our industrial collaborators, Westpoint Veterinary Group, AHDB, Volac, MSD and Arla for detailed discussion of project milestones and initial data collection.
Year(s) Of Engagement Activity 2018
 
Description Fellowship press release 
Form Of Engagement Activity A press release, press conference or response to a media enquiry/interview
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Media (as a channel to the public)
Results and Impact Press release written by Holly Kernot from The Veterinary Times highlighting the success of the researchers in winning the fellowship and the potential merits of the project
Year(s) Of Engagement Activity 2018
URL https://www.vettimes.co.uk/news/doctors-invest-600k-in-novel-brd-detection/
 
Description Meeting with international Advisory Board members 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Organised and held initial meeting with our European collaborators in Baja, Hungary - IRT Diagnose KG (Austria), ROBDOK Kft (Hungary), Siculovet SRL (Serbia), Vetservis (Romania) - for detailed discussion of project milestones and potential recruitment of European farms.
Year(s) Of Engagement Activity 2018
 
Description Poster presentation at BMVA symposium "Visual Image Interpretation in Humans and Machines" 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact This one-day meeting (10th April 2019) considered the issue in human and machine vision discussing how artificial neural net-works might be augmented with more biologically plausible features with the aim of making them more robust, and alternatives to neural network models and how their performance compares to the state of the art and human vision.
Year(s) Of Engagement Activity 2019
URL http://www.bmva.org/meetings
 
Description Poster presentations at British Cattle Vet Association Congress 2019 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Two posters were presented at BCVA 2019, which elicited conversations with researchers in veterinary medicine regarding automation on farms (in particular cattle farms).
Year(s) Of Engagement Activity 2019
URL https://www.bcva.org.uk/cpd/Congress-Programme
 
Description Talk at Blagdon Primary School (Blagdon) 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Schools
Results and Impact Approx. 50 pupils attended for a talk on our group's research, which sparked several questions and discussion afterwards.
Year(s) Of Engagement Activity 2018
 
Description Visiting Houghton Bloodstock near Newmarket to set up a collaborative project 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Professional Practitioners
Results and Impact We have visited the Houghton Bloodstock stud near Newmarket to discuss a collaboration between Rossdales Veterinary Surgeons, Houghton Bloodstock and University of Bristol on automatic disease detection in newborn foals using the sensors and software developed through the Fellowship. Two vets and one of the owners of the bloodstock was present, who were very interested in the project and promised resources (on-site assistant, veterinary records, access to horses). After successfully securing industrial partners, we are submitting an EPSRC IAA Exploratory Award application to explore disease monitoring in foals and generate impact in Newmarket, the global centre of thoroughbred horse racing, the largest racehorse breeding centre in the country and most importantly, a key global centre for horse health.
Year(s) Of Engagement Activity 2019
 
Description Visits to three ranches in Austria and meeting with a collaborating vet in Hungary to discuss automatic monitoring of horses organised by industry partner IRT Diagnose KG 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact Flying P Ranch
Near Mannersdorf

Visited the Flying P Ranch to attend a thermal camera examination of a lame horse with a badly fitted horseshoe and assess the standards of horse keeping in Austria (approx. 40 horses).

Western Star
Near Wiener Neustadt

Visited Western Star, the biggest quarterback horse stable in Austria (approx. 150 horses) and one of the centres of western riding in Europe. The ranch regularly holds international competitions with horses flew in from all over the world. During a meeting with owner and founder, Helmut Schulz, we were advised that the most effective way to generate impact with new technological solutions to the racehorse industry is via organising workshops for trainers. Also, automatic data collection of horses could result in confidentiality problems, something we need to be aware of and actively seek solutions for. We also learned that most diagnostics focus on horse limbs and there is potential in conducting research on back and hip problems - issues which often rise from inappropriate riding.

Bauer Reitstall
Near Pöttsching

Following an extremely informative meeting with Eike Both, a horse trainer with veterinary qualifications, we learned that horse flu is a very common problem in the area. We discussed monitoring the disease with sound by placing several microphones around the stable and triangulating coughs. This could be a cost-effective way to detect early signs of disease even in small stables. Strangles (Streptococcus equi infection) is also an issue and we discussed potential ways of detecting the disease automatically. An opportunistic outcome of the meeting was a discussion on automatic monitoring of puppies, which often end up with respiratory problems after transportation - something to develop further with dog expert colleagues in Langford.

Meeting with József Hegedus
Baja

József Hegedus is a veterinary surgeon in the Vojvodina region of Serbia (North Serbia), who is already an international advisor on our EPSRC fellowship. This meeting was organised in order to discuss common horse problems that could be monitored via automatic imaging. József highlighted that he deals with a lot of behavioural issues with horses, partially coming from transportation stress. He mentioned that several breeders now also focus on anxiety and try to breed horses that handle being left alone in a box better. He recommended looking at behaviours such as biting handles as monitoring these would be highly informative for vets. He also knows stables which could be recruited for research in the future.
Year(s) Of Engagement Activity 2019