A key player in a revolutionary wave of automated healthcare, Vet-AI operates through its consumer-facing platform Joii, a petcare app.
Using online symptom checkers and phone-based consultancies, the company aims to keep pets and their owners out of veterinary practices aiming to reduce both practitioner burn-out rates – an issue facing the industry – and the financial burden on pet owners.
The challenge
In a move towards using artificial intelligence in symptom diagnosis, Vet-AI initially chose to focus on a single common issue: the recognition of gait abnormalities amongst dogs. Could the business create a tool that would triage an animal exhibiting lameness, establish the severity of that lameness and possibly prevent unnecessary visits to the veterinary practice?
The project team faced numerous challenges including the gathering and labelling of sufficient data to create an accurate model, and the programming of the model itself. This project needed to demonstrate robust scientific methods before it could progress towards commercial application.
The solution
Through a Knowledge Transfer Partnership (KTP) with the University of Leeds, the Vet-AI team collaborated with academic specialists including David Hogg, Professor in Artificial Intelligence and an internationally acclaimed expert in computer vision, Dr Derek Magee, an expert in computer vision and Dr Robert Cooper, who initially joined as the KTP Associate and is now senior AI developer at Vet-AI.
David and Derek used their expertise in three-dimensional geometric modelling for tracking flexible structures to help identify healthy and abnormal gaits among dogs. The Vet-AI team collaborated with charities including The Dog's Trust to amass sufficient footage for analysis. They also employed canine biomechanics experts to help catalogue and label the data, describing the extent and severity of an animal's lameness.
Using this variety of complex data and under the University's supervision, Dr Robert Cooper created an algorithm using convolutional neural networks that would form the basis of the petcare app, Joii.
The impacts
Collaborating with the University has enabled Vet-AI to create a working model that identifies and analyses problematic dog gaits with an impressive 70% accuracy to date, reducing the need for veterinary intervention.
As CEO Paul Hallett points out, ‘the same visual computing methodology can be applied to other common animal issues, including the recognition of skin lesions and pain experienced by cats’.
Vet-AI's partnership with the University means that findings have been tested rigorously in an academic environment and this rigour has helped Vet-AI attract over £12m investment since its inception.
Vet-AI looks forward to additional future collaborations with the University and following this ground-breaking project they have secured two more KTPs to further drive R&D and innovation in pet health.
Video transcript: University of Leeds KTP with Vet-AI.
To discuss how working with us on a KTP could help your business, email the Knowledge Transfer Partnership team via ktp@leeds.ac.uk.