Improving Diagnosis of Canine Leptospirosis: Development of a Machine Learning Tool using UK and Canadian Data

Canine leptospirosis is a climate-sensitive, bacterial, infectious disease of global importance. Despite being vaccine-preventable, leptospirosis is an important cause of morbidity and mortality in dogs. Across North America and the United Kingdom (UK), the prevalence of canine leptospirosis ranges between 4.8-14%, with evidence for increasing prevalence over time. Leptospirosis can be difficult or slow to diagnose due to non-specific clinical signs and need to test blood or urine using laboratory-based diagnostics. Early diagnosis is important for initiation of treatment and improvement of outcomes. Thus, there is a need for clinical tools that can be used to assist with early detection of leptospirosis. Machine learning methods have been shown to aid in early and accurate diagnosis for a variety of diseases. We propose to develop a validated machine learning model trained on canine clinicopathologic, signalment, and environmental information using a large, geographically and clinically representative canine population using data from the United Kingdom and Canada. This model will extend prior research by including environmental data, such as season, temperature, or precipitation, which may improve its predictive ability. This model can be applied by veterinarians to improve their ability to diagnose leptospirosis and implement earlier treatment in Canada and the UK.

Faculty Supervisor:

Lauren Grant

Student:

Partner:

University of Liverpool

Discipline:

Life Sciences

Sector:

Health and Related Sciences & Technology

University:

University of Guelph

Program: