Evaluation of infrared thermography body temperature and collar-mounted accelerometer and acoustic technology for predicting time of ovulation of cows in a pasture-based system

This study was conducted to test the hypothesis that the specificity of infrared thermography (IRT) in detecting cows about to ovulate could be improved using different body parts that are less likely to be contaminated by fecal matter. In addition, the combined activity and rumination data captured by accelerometers were evaluated to provide a more accurate indication of ovulation than the activity and rumination data alone. Thermal images of 30 cows were captured for different body areas (eye, ear, muzzle, and vulva) twice daily after AM and PM milking sessions during the entire experimental period. Milk progesterone data and insemination records were used to determine the date of ovulation. Cows were fitted with SCR heat and rumination long-distance tags (SCR HR LD) for 1 month. Activity- and rumination-based estrus alerts were initially identified using default threshold values set by the manufacturer; however, a range of thresholds was also created and tested for both activity and rumination to determine the potential for higher levels of accuracy of ovulation detection. Visual assessment of mounting indicators resulted in 75% sensitivity (Se), 100% specificity (Sp), and 100% positive predictive value (PPV). Overall, IRT showed poor performance for detecting cows about to ovulate. Vulval temperature resulted in the greatest (80%) Sp but the poorest (21%) Se compared with the IRT temperatures of other body areas. The SCR HR LD tags default threshold value resulted in 78% Se, 57% Sp, and 70% PPV. Lowering the activity threshold from the default value improved the sensitivity but created a large number of false positives, which resulted in a decrease in specificity. Lowering the activity threshold to 20 resulted in a detection performance of 80% Se, 94% Sp, and 67% PPV, whereas the rumination levels achieved 35% Se, 69% Sp, and 14% PPV. The area under the curve for the activity level, rumination level, and the combined measures of activity and rumination levels were 0.82, 0.54, and 0.75, respectively. Alerts generated by SCR HR LD tags based on a lower activity threshold level had high sensitivity and may be able to detect a high proportion of cows in ovulatory periods in pasture-based system; however, the specificities and positive predictive value were lower than the visual assessment of mounting indicators.