Artificial intelligence in smartphone video analysis for equine asthma diagnostic support.

Abstract

Background: Equine asthma is a prevalent respiratory disease that negatively impacts horses' health and athletic performance. Traditional diagnostic methods are invasive and require specialised equipment. There is a need for a non-invasive, cost-effective screening tool that can be used by veterinarians and horse handlers in ambulatory settings.

Objectives: To assess the willingness of veterinarians and horse handlers to adopt such a tool (Questionnaire 1) and the challenges associated with visually recognising equine asthma (Questionnaire 2) and to develop EquiBreathe, an artificial intelligence (AI)-powered, non-invasive diagnostic tool designed to enhance equine asthma detection.

Study design: Cross sectional survey and AI model development.

Methods: Two Google Forms questionnaires were distributed. Video recordings of 23 horses (12 diagnosed with asthma and 11 healthy controls) were collected, focusing on nostril and abdominal movements. AI models were trained using feature engineering and image subtraction techniques.

Results: Questionnaire 1 was completed by 18 veterinarians, 24 veterinary students and 121 horse handlers, while Questionnaire 2 involved 10 veterinarians, 23 students and 13 handlers. Respondents showed strong interest in the tool, emphasising its potential to improve communication and diagnostic precision (Questionnaire 1). However, relying solely on visual assessment for asthma detection proved difficult for veterinarians (Questionnaire 2), underscoring the value of AI support. The best-performing AI model achieved 89% accuracy in distinguishing asthmatic from healthy horses using nostril data.

Conclusions: The study demonstrated the need for a field-friendly diagnostic solution. EquiBreathe was shown to have promising potential as a non-invasive, cost-effective screening tool.