Assessment of Neck Mounted Accelerometer Technology for Detecting Oestrus in Pasture-Based Dairy Cows

Accurate visual detection of oestrus in dairy cows is time consuming and labour intensive (Hockey et al 2010). Cow mounted technology has been developed to detect oestrus, but there is limited research regarding the use of this technology for cows in pasture-based systems. This study was conducted to test the hypothesis that a commercially available accelerometer device would allow for detection of oestrus as determined by milk progesterone levels to a higher level of accuracy than visual detection aided with mounting detectors (Estrotect^TM, Rockway, Inc., US) placed on the tail head of pasture-based dairy cows. Thirty cows were fitted with accelerometers (SCR HR LD neck collars, Netanya, Israel) for a 1 month period. With the same collar, the devices continuously profiled the changes in individual cow activity and rumination for 2 h time blocks. Using a mathematical algorithm, a weighted activity and rumination index was calculated by the software that expressed the momentary deviation of the activity and rumination from the average activity in the same time period during the past 7 days. Activity based oestrus alerts were initially identified using default threshold values set by the manufacturer (presented here as combined activity and rumination level), however, a range of thresholds were also tested for activity and rumination level independently to determine their potential in relation to the accuracy of oestrus detection. The sensitivity and specificity were calculated at several thresholds and the overall utility of the method was assessed through construction of a receiver operator characteristic (ROC) curve. After the points were connected with lines, the area under the curve (AUC) was calculated using a simple trapezium method. The AUC (0 < AUC < 1) is commonly used as a measure of test performance, with AUC = 0.5 corresponding to a test performance no better than chance. Milk progesterone data and insemination records were used to determine the 'true' oestrus date. Visual assessment of the Estrotect devices resulted in 26% sensitivity (Se), 100% specificity (Sp) and 100% positive predictive value (PPV). The activity level performed better than the rumination level. At a threshold of less than 20, activity level had a detection performance of 80% Se, 94% Sp and 67% PPV, while the rumination level achieved 35% Se, 69% Sp and 14% PPV. The AUC for activity level, rumination level and the combined activity and rumination level were 0.82, 0.54 and 0.75 respectively which indicates that the activity level provided a more accurate indication of likely oestrus events than the rumination level or the combined rumination and activity level (Figure 1).

The results gathered in this study indicate that the SCR neck collars can achieve higher performances compared to Estrotect in detecting cows in oestrus. The activity level had the highest sensitivity and may be a useful aid in the detection of a high proportion of cows in oestrus in pasture-based systems. However, the specificities and positive predictive value may be lower than the visual assessment of mounting indicators. This suggests that a high proportion of oestrus cows will be detected through the use of the SCR neck collars, but that these alerts will require filtering either through new algorithms or through the combination of other technologies to remove the false positives.

Hockey, C.D., Morton, J.M., Norman, S.T. and McGowan, M.R. (2010). Reprod. Domest. Anim. 45, e107.

This work was supported by the Future Dairy project