Mastitis is one of the most severe diseases in dairy production worldwide, with a high prevalence in machine milking cows causing significant economic losses due to decreased milk production, deteriorated milk quality, increased treatment costs, reduced longevity and increased culling rate (Halasa et al. Reference Halasa, Nielen, De Roos, Van Hoorne, de Jong, Lam, Werven and Hogeveen2009). Infrared thermography (IRT) is a non-invasive and non-contact technique for measuring surface heat emitted as infrared radiation and generates pictorial images that represent the amount of heat with different colours (Kunc et al. Reference Kunc, Knizkova, Prikryl and Maloun2007). In recent years, extensive researches have been conducted to confirm the ability and feasibility of IRT to detect mastitis in dairy cows, sheep, goats and camels. Studies showed that changes in the udder skin surface temperature in the range of 1·0–2·35 °C were detected in subclinical/clinical mastitis dairy cows (Scott et al. Reference Scott, Schaefer, Tong and Lacasse2000; Polat et al. Reference Polat, Colak, Cengiz, Yanmaz, Oral, Bastan, Kaya and Hayirli2010). Among these studies, most were conducted pre-milking (Polat et al. Reference Polat, Colak, Cengiz, Yanmaz, Oral, Bastan, Kaya and Hayirli2010), some during or soon post-milking (Berry et al. Reference Berry, Kennedy, Scott, Kyle and Schaefer2003), and very few some time post-milking. However, udder skin surface temperature variation between pre- and post- milking and optimal time to capture infrared images of udders by IRT to detect mastitis is not yet determined. More information regarding the optimal time for collecting images around milking could improve the sensitivity and specificity of IRT for detecting mastitis. Therefore, this study aimed to investigate the udder skin surface temperature variation pre- and post-milking to determine the optimal time for collecting images for improving the sensitivity and specificity of IRT in dairy cows' mastitis detections.
Materials and methods
The experiment was performed in accordance with guidelines of the Beijing Animal Ethics Committee and approved by the Chinese Academy of Agricultural Sciences Animal Care and Use Committee. One hundred and two lactating Chinese Holstein dairy cows (1st parity, n = 57; 2nd parity, n = 12; 3rd parity, n = 33) were used in the trial, with average of 76 ± 67 d in milk and 13·5 ± 4·7 kg of milk yield at morning milking (Mean ± sd).
Infrared images of left and right hind quarters from the caudal view were captured pre-milking and post-milking. Meanwhile, the milk yield of each cow at the morning milking was recorded. In a quarter thermograph image, the maximum value of an area of a circle with the diameter of 40 pixels above the teat base was measured with the help of the camera software (IRBIS 3 Standard, InfraTec GmbH) and used in the data analysis. Detailed information on animal preparation, experimental procedures, udder infrared images processing and udder skin surface temperature measurement are reported in the Supplementary File as online Supplementary Materials.
All statistical analyses were performed with SAS 9·2 software (SAS Institute, Cary, NC). A comparison was made between the left and right hind quarter skin surface temperature pre- and post-milking, and the data was analysed by a paired two-sample test. The comparison of hind quarter skin surface temperature between pre-milking and post-milking was analysed by the paired t-test. The correlation of milk yield and hind quarter skin surface temperature was analysed with PROC CORR. The effect of milk yield on the hind quarter skin surface temperature was analysed by one-way ANOVA. Statistical significance was assumed at P < 0·05.
Results and discussion
No significant difference was observed between left and right hind quarter skin surface temperature pre-milking and post-milking and the range of temperature variation between left and right hind quarter was less than 1·0°C (P > 0·05; Table 1). The bilateral symmetry characteristic of skin surface temperature shown in this study by left and right hind quarters was similar to the result of Saraiva Martins et al. (Reference Saraiva Martins, Paim, Cardoso, Lima Dallago, de Melo, Louvandini and McManus2013) and Metzner et al. (Reference Metzner, Sauter-Louis, Seemueller, Petzl and Klee2014), who reported that there were no substantial differences in udder surface temperature between left and right hind quarters. This may be related to the bilateral symmetry characteristic of the udder tissue structure and blood circulation system. Since the udder skin surface temperature reflects the tissue metabolic activity and blood circulation status, the left and right quarter show a similar temperature distribution.
Table 1. The left and right hind quarter skin surface temperature distribution pre- and post-milking The left and right hind quarter skin surface temperature measured 20–30 min pre-milking and 25–30 min post-milking from 102 Holstein dairy cows at the morning milking in one day

LQSST, left quarter skin surface temperature; RQSST, right quarter skin surface temperature; TD(left-right): temperature difference between left and right hind quarter; TD(post-pre milking): temperature difference between post-milking and pre-milking. Within columns and rows, values with different superscript letters are significantly different (P < 0·05).
Since the left and right quarter skin surface temperatures were characterised by symmetrical distribution, their averaged values were calculated as the whole hind quarter skin surface temperature. In this study, we divided the milk yield at the morning milking into three levels: 0–10 kg, 10–15 kg, and 15–20 kg and analysed the effect of milk yield on hind quarter skin surface temperature. The results showed that the hind quarter skin surface temperature pre-milking was not correlated with the milk yield at morning milking (P = 0·26). The milk yield exerted no significant effect on quarter skin surface temperature pre-milking (P = 0·10, Table 2). However, quarter skin temperature pre-milking manifested an upward tendency along with the augmentation of milk yield. The average hind quarter skin surface temperature variation of cows with different milk yield was less than 0·7°C. Schmidt et al. (Reference Schmidt, Bowers, Dickerson, Graves and Willard2004) found that cows with a high milk production had higher udder skin temperatures pre-milking than did the low producing cows. These results may be related to the gland cell metabolic activity, blood flow velocity, and the mammary gland volume. Before the milking process, milk has accumulated in the secretory tissue and cistern of the mammary gland, blood flow may be suppressed and the metabolic activity of the mammary secretory cells is reduced. Therefore, udder skin surface temperature pre-milking was not influenced by milk yield at morning milking.
Table 2. Effect of milk yield on hind quarter skin surface temperature pre- and post-milking The left and right hind quarter skin surface temperature of dairy cows with different milk yield at the morning milking

Data are presented as means ± sem. HQSST pre-milking: hind quarter skin surface temperature pre-milking; HQSST post-milking: hind quarter skin surface temperature post-milking
a,b: Means within a row with different superscripts differ (P < 0·05).
On the contrary, the udder skin surface temperature post-milking was positively correlated with the milk yield (R = 0·31, P < 0·01). The udder skin surface temperature of cows with a high milk yield was 0·98°C higher than that in cows with a low milk yield (P = 0·01, Table 2). This suggests that the udder skin surface temperature post-milking was influenced by milk yield at morning milking. Previous studies showed that the machine milking process had a great influence on udder skin surface temperature and may cause an instant rise of more than 1·0°C (range from 0·8–2·65°C) compared with its pre-milking values (Kunc et al. Reference Kunc, Knizkova, Prikryl and Maloun2007; Vegricht et al. Reference Vegricht, Machalek, Ambroz, Brehme and Rose2007). Although mechanical milking can cause circulatory changes in teat tissue fluids, inducing an increase in teat skin temperature (Hamann et al. Reference Hamann, Burvenich, Mayntz, Osteras and Haider1994), the changes lasted only a very short time after milking (Isaksson & Lind, Reference Isaksson and Lind1994). In our study, udder infrared images were obtained 20 to 35 min post-milking in order to avoid the influence of machine milking on udder skin surface temperature. We suggest that, at approximately 20 to 35 min post-milking, the mammary gland cells and vascular system were free of milk accumulation, and cell metabolic activity and blood flow returned to maximal values. Therefore, cows with an elevated mammary metabolic activity had a higher quarter skin surface temperature than did the cows with low metabolic rates. In this study, the temperature difference of the left and right hind quarter post-milking was 1·29 and 1·31°C higher than that before milking, which was similar to the temperature changes caused by an intramammary infection (Samara et al. Reference Samara, Ayadi and Aljumaah2014), indicating that the udder infrared images captured pre- and post- milking could not be used simultaneously. In addition, the udder skin surface temperature of cows with a high milk yield was 0·98°C higher than that in cows with a low milk yield, which was nearly equal to the temperature changes induced by subclinical mastitis (Scott et al. Reference Scott, Schaefer, Tong and Lacasse2000; Polat et al. Reference Polat, Colak, Cengiz, Yanmaz, Oral, Bastan, Kaya and Hayirli2010). Therefore, the sensitivity and specificity of IRT in mastitis detection by capturing udder infrared images post-milking may be influenced by milk yield. Considering that udder skin surface temperature pre-milking was not influenced by milk yield, it may be better to measure udder skin surface temperature pre-milking in order to avoid the influence of physiological parameters and improve the sensitivity and specificity of IRT to detect mastitis in dairy cows.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022029918000213
The study was financially supported by Beijing Dairy Industry Innovation Team Project (BAIC06-2017), National Key Research and Development Program of China (2016YFD0500507, 2017YFD0502003) and The Agricultural Science and Technology Innovation Program (ASTIP-IAS07).