AI-powered acoustic surveillance for early detection of calf respiratory disease

Abstract Effective management of Bovine Respiratory Disease Complex (BRDC) requires timely, non-invasive diagnostic tools to protect calf health and welfare. Among early clinical signs, coughing stands out as both frequent and informative. To explore its potential for early BRDC detection, we deployed an artificial intelligence (AI)-driven acoustic monitoring system that recorded over 2,730 hours of audio during a 30-day period. Four experimental pens, each housing seven calves and stratified by infection status and antibiotic treatment, were equipped with a dedicated microphone to enable targeted acoustic surveillance. This configuration enabled pen-specific detection of cough events, which were subsequently classified using an AI HuBERT-based model trained on 1,045 labelled clips. The classifier achieved 92% accuracy. Temporal patterns in cough frequency aligned with infection dynamics, treatment responses, and circadian patterns. Notably, AI-detected coughs consistently preceded clinical scores by 1–2 days, confirming the system’s sensitivity to early respiratory disorders. These findings support the use of acoustic surveillance as a valid, scalable, and autonomous tool for continuous monitoring and early warning of respiratory diseases in calves. Implications This study demonstrates that AI-powered acoustic monitoring enables real-time, non-invasive detection of coughs in calves for early warning of respiratory diseases, outperforming traditional veterinarian clinical scoring by 1–2 days. Its high accuracy and sensitivity to respiratory infection dynamics and treatment effects position it as a scalable tool for precision livestock farming.