Using audio-visual monitoring to evaluate immune and behavioral indicators to lipopolysaccharide challenge in beef cattle

Abstract

This experiment investigated the integration of camera and microphone monitoring with AI technologies to assess immune responses, activity patterns, and respiratory acoustic changes in beef cattle using lipopolysaccharide (LPS) administration as a model. Twelve Angus steers [age = 15 ± 1 mo; body weight (BW) = 399 ± 6 kg] were housed in four feedlot pens (3 steers/pen). On day 0, steers were ranked by initial BW, and assigned to one of two groups in a crossover design, with treatments initiated at hour 0: 1) an intravenous bolus dose of bacterial LPS (0.5 μg/kg of BW; n = 12), or 2) 5 mL intravenous injection of 0.9% sterile saline (CON; n = 12). The experimental period lasted 9 days (days -3 to 6) with 20 days of washout between periods. Video footage was recorded, and animal positions were tracked using a deep learning-based object detection algorithm (YOLOv11). Video frames were used to generate activity heatmaps for three consecutive one-hour intervals: h 0-1, 1-2, and 2-3. Microphones mounted on each animal's halter were used to measure differences in breathing patterns. Steers rectal temperature was assessed, and blood samples were collected at hours 0, 4, 8, 24, 48, 72, and 96 relative to treatment administration. LPS induced a neuroendocrine response by increasing plasma cortisol concentrations (P ≤ 0.04) at 4, 8, and 12 h, non-esterified fatty acid (NEFA) at 8, 12, and 24 h, while reducing (P < 0.01) circulating beta-hydroxybutyrate (BHBA) at 4 and 8 hours vs. CON group. LPS steers displayed an increase (P < 0.01) in rectal temperature at 4 and 12 h, and a reduction (P < 0.01) in intake at 24, 48, and 72 h post-challenge vs. CON cohorts. The YOLOv11 identified reduced pen exploration and increased clustering behavior among LPS steers 1.5 h post-challenge. Spatial analysis of steer movement further confirmed that LPS steers exhibited increased resting behavior and avoided the feeding bunk throughout the 3-hour window. In parallel, Mel spectrogram analysis of breathing sounds 2 hours after LPS revealed an altered respiratory pattern in LPS steers, characterized by short, sharp, and high-pitched sounds suggestive of labored breathing. Collectively, LPS administration elicited behavioral and physiological changes, which were detected using integrated AI-driven video and audio monitoring systems. These findings suggest that these technologies may offer a promising approach for real-time assessment of cattle behavior and respiratory dynamics in cattle.