162 Evaluating the Potential of a Thermal Imaging System to Identify Subclinical Mastitis in Dairy Cattle

Abstract

Abstract The objective was to evaluate the potential for the use of a thermal imaging system for the identification of subclinical mastitis in dairy cows. Multiparous Holstein dairy cows (n = 131) within 150 days in milk (DIM; range 65 to 155) were enrolled in the study at a commercial dairy farm in central Texas. Milk samples (10 mL) were collected from each quarter separately and evaluated for somatic cell counts (SCC; n*1000), fat (%), protein (%), lactose (%), SNF (%), BHBA (mM/L), acetone (mM/L), and milk urea nitrogen (MUN; mg/dL). A minimum of four thermal images per cow representing all four quarters were captured using a handheld infrared camera (Fluke TiS 75+) from approximately one-meter distance from the cow. All images were taken inside the milking parlor and after the morning milking during February 2023 to minimize the impact of daily ambient temperatures. Skin temperatures from each quarter (maximum, minimum, average) were extracted using Fluke Connect software. A total of 524 quarter-level milk samples and corresponding udder images were obtained, and 10 images were discarded from analyses due to the lack of appropriate resolution and focus. To prevent collinearity, a correlation matrix was developed using PROC CORR in SAS and variables with significant correlations were removed from the model. Subclinical mastitis was defined at the quarter level when SCC > 250,000. Statistical analyses were performed using mixed linear regression models (PROC MIXED) with cow and DIM as random effects in the model, and logistic regression model (PROC LOGISTIC) was fitted to evaluate the probability of the disease based on temperature and milk variables. Maximum and average udder temperatures did not demonstrate a significant relationship with subclinical mastitis status whereas minimum udder temperature was greater in sick quarters (29.6 ± 0.52 vs 27.5 ± 0.45; P = 0.003). Increased minimum udder temperature was also associated with BHBA (P < 0.0001), acetone (P < 0.0004), and MUN (P < 0.0007) concentrations in milk. Cows with subclinical mastitis had decreased BHBA (0.14 ± 0.037 vs 0.28 ± 0.031; P = 0.0036) and acetone (0.017 ± 0.003 vs 0.044 ± 0.003; P < 0.001). Odds of subclinical mastitis were associated with MUN values in milk (OR 0.94; CI 0.89 - 0.98) but not with any udder surface temperature variables. We conclude that an increase in the minimum temperature of the udder could be used as an indicator of subclinical mastitis but studies with larger datasets and using advanced image analysis techniques are warranted to validate these results.