Is the application of thermography for bovine disease diagnostics a suitable approach for developing a machine learning algorithm?

Abstract- Contemporary technological solutions provide significant opportunities not only to optimize dairy production on farms, but also to monitor cows physiological parameters and the overall health status of the herd. One of the most advanced animal monitoring approaches is the application of thermography. This non-invasive method enables the detection of early changes in surface temperature, particularly for identifying inflammatory processes in the mammary gland and limbs. The integration of infrared thermography (IRT)technology into the milking parlor, together with the development of machine learning algorithms (MLA), could facilitate the monitoring and early detection of limb disorders and initial signs of mastitis in cows. Early identification of subclinical disease, before the onset of clinical symptoms such as lameness, would allow timely preventive interventions, thereby reducing treatment expenses and economic losses. Lameness is one of the most prevalent and economically significant health problems affecting dairy farms worldwide. It has been shown that 55% of lactations are associated with lameness-related health issues, while 15% are linked to mastitis or uterine infections. The aim of this study was to evaluate the effectiveness of IRT method in the diagnosis of bovine limb and udder diseases, with the goal of establishing a foundation for the development of MLA. For this study, a FLIR T640 thermal camera (FLIR Systems, USA) was used: Upon entering the milking parlor, thermographic images of the front and hind limbs and of the udder (with all teats visible) were captured (ambient temperature 15°C). Image analysis was performed using the Flir Tools software. The cows milk was tested by express diagnostic method before milking using the CMT (California Mastitis Test). During milking, milk samples were collected for the determination of lactose, fat, urea, protein concentrations, and somatic cell count. After milking, clinical examinations of all limbs were conducted to assess claw pathologies. Data analysis revealed that in the region of the articulationesinterphalangeaedistales, the surface temperature of clinically healthy front limbs was 18.06 +- 1.8 C whereas limbs with confirmed pathology exhibited a temperature of 27.31 +- 3.8 C - an increase of 9.25 C (p 1 < 0.001 ). In the hind limbs, the surface temperature of clinically healthy limbs was 19.58 = 2.62 C while limbs with pathology reached 29.77 +- 2.5 C - an increase of 10.19 C (p < 0.001). Analysis of teat surface temperature showed that healthy teats exhibited a temperature of 25.82 +- 2.43 C whereas showing signs of subclinical mastitis (confirmed by CMT) reached 30.42 +- 0.92 C an increase of 4.6 C (p < 0.001). Spearman's correlation analysis indicated that clinically confirmed signs of inflammation in the hind limbs were associated with increased teat temperatures left front (LF) (rs = 0.21, p<0.01), right front (RF) (rs = 0.22 p < 0.01), left rear (LR) (rs = 0.22, p < 0.01), and right rear (RR) (rs = 0.24 , p < 0.001). However, clinically confirmed signs of inflammation in the for climbs were not associated with the temperature of all four teats (p > 0.05) . Additionally, increased teat surface temperature was negatively associated with milk protein and urea content (rs = -0.18 to -0.28, p < 0.05 ), irrespective of temperature changes detected in the front and hind limbs. The results of this dataset for the development of machine learning algorithms capable of identifying associations between limb and udder diseases in dairy cows and to predict their impact on herd health.