Evaluating progesterone profiles to improve automated oestrus detection
•Download as PPTX, PDF•
2 likes•412 views
How well are automated oestrus detection system in finding cows in heat? A presentation at the EAAP/EU-PLF Conference, 2014, in Copenhagen, Denmark
Recommended
Recommended
More Related Content
What's hot
What's hot (15)
Similar to Evaluating progesterone profiles to improve automated oestrus detection
Similar to Evaluating progesterone profiles to improve automated oestrus detection (20)
Recently uploaded
Recently uploaded (20)
Evaluating progesterone profiles to improve automated oestrus detection
- 1. Evaluating progesterone profiles to improve automated oestrus detection Claudia Kamphuis, Wageningen University Kirsten Huijps, CRV Henk Hogeveen, Wageningen and Utrecht University
- 2. - First insemination between 40-70 DIM is optimal* - oestrus detection is a key driver - Challenges of detection - Time consuming - Error prone - Increased herd sizes The importance of oestrus detection *Inchaisri et al., 2012
- 3. - 20% of Dutch dairy farms have automated systems* - Appears to be a success story - 20% of Dutch dairy farms have automated systems* - Appears to be a success story when it all works - Sensitivities 80-90% with specificities > 90%** - No technical / human errors Adoption automated oestrus detection *Huijps, 2014, CRV, personal communication*Huijps, 2014, CRV, personal communication **Rutten et al., 2013
- 4. - Normal cycle takes 21 days - Does progesterone affect oestrus behaviour - Affect automated oestrus detection? Progesterone and oestrus Days Progesterone(ng/ml) Behavioural changes
- 5. Aims of this study: gain insight in - Performance of oestrus detection in the field - Timing of oestrus alerts - Use of alerts by farmers - Effect of combining oestrus alerts on performance - Effect of progesterone profiles on oestrus detection
- 6. Materials and Methods - 31 cows, 40-70 DIM, not inseminated Farm A (450) Farm B (AMS; 250) Milk samples for 24 days Morning milkings Residual milk 12 cows First milkings Whole milk 19 cows Oestrus alerts System A: 12 cows System B: 12 cows System B: 8 cows System C: 19 cows Oestrus observations Farm Staff Farm Staff Average DIM at start 44 53 Average Parity 5.5 2.4
- 7. Hormonost-Microlab Farmertest, Biolab, Unterschleissheim, Germany Progesterone profiles from milk samples - Commercial on-farm kit - Analyses 3x a week - Including forgone 1 / 2 days - Profiles created - Visual assessment of heat - According to manual - Gold standard
- 8. Results: heats, observations and alerts - Based on Progesterone (P): 30 heats from 30 cows Farm Staff System A B C Heat observed/alerts generated 15 14 12 31 Heat alerts on day with P-heat 3 9 Heat alerts on day with P-heat +/- 1 day 9 17 False positive observations / alerts 6 4 5 18
- 9. Results: timing alerts and observations 0 1 2 3 4 5 6 -23 -13 -3 7 17 Numberof alerts/observations days around day of P-heat (day = 0) System A
- 10. Results: timing alerts and observations 0 1 2 3 4 5 6 -23 -13 -3 7 17 Numberof alerts/observations days around day of P4heat (day = 0) System A System B
- 11. Results: timing alerts and observations 0 1 2 3 4 5 6 -23 -13 -3 7 17 Numberof alerts/observations days around day of P4heat (day = 0) System A System B System C
- 12. 0 1 2 3 4 5 6 -23 -13 -3 7 17 Numberof alerts/observations days around day of P4heat (day = 0) Farm staff System A System B System C Results: timing alerts and observations 84% of all alerts 87% of all observations are 3 days around P-heat False or True?
- 13. Results: combining detection systems Using a 1 day time window around a P-heat Farm A System A: 5 out of 12 P-heats (42%) System B: 3 out of 12 P-heats (25%) One P-heat additionally detected Farm B System B: 2 out of 18 P-heats (11%) System C: 9 out of 18 P-heats (50%) No additionally P-heat detected
- 14. Results: effect of progesterone profiles 0 5 10 15 20 25 30 -23 -20 -17 -14 -11 -8 -5 -2 1 4 7 10 13 16 19 22 Progesteronelevel(ng/mL) Days around P-heat (day = 0) Average P levels
- 15. Results: effect of progesterone profiles 0 5 10 15 20 25 30 -23 -20 -17 -14 -11 -8 -5 -2 1 4 7 10 13 16 19 22 Progesteronelevel(ng/mL) Days around P-heat (day = 0) Average P levels Detected P-heats
- 16. Results: effect of progesterone profiles 0 5 10 15 20 25 30 -23 -20 -17 -14 -11 -8 -5 -2 1 4 7 10 13 16 19 22 Progesteronelevel(ng/mL) Days around P-heat (day = 0) Average P levels Not detected P-heats Detected P-heats
- 17. Conclusions * Rutten et al., 2012; Kamphuis et al., 2012 - All 3 systems performed less than expected - Expected: 80%*; Found: 25-50% - Farm staff missed 48% of true positive alerts - Not checked alerts / behavioural changes already passed - Most alerts and observations around 3 days of P-heat - Progesterone profiles did not differ between (non)detected cows
- 18. Take Home Message - Farmers miss correctly identified oestrus's - Progesterone does not affect oestrus behaviour - Confirm with larger numbers - Successful inseminations as gold standard
- 19. Acknowledgements - Hands-on - Farmers - Farm staff - Marije Popta and Gea Miedema (Van Hall-Larenstein, Leeuwarden) - Funding
Editor's Notes
- Thank chairman and audience. This presentation contains three parts: the first addresses oestrus, oestrus detection, progesterone and aims of this study. The second part covers the materials and methods used to answer these aims. The last part covers the results and the take home message from this research.
- Why is it so important that we know when a cow is oestrus? Simple; we want milk. To get milk, we need to cow to get a calf. We know that a calving interval of one calf per year is the most optimum from an economic point of view. To reach this interval, the cow has to be inseminated when she is 40 to 70 days in milk. You only want to inseminate cows when they are in oestrus, ready to conceive, thus oestrus detection is a key driver for a one-year calving interval. Normally, farmers look for oestrus behaviour to find cows in heat, with mounting or being mounted being very clear signs of oestrus. However, there are challenges. Firstly, observing for oestrus behavour is time consuming, requires attention and experience. It’s error-prone as you have to know the cow that expresses oestrus behaviour. This may be not a problem in small herds, but with the increased herd sizes this is challenging. Because who can tell who’s in oetrus now? To help farmers with this challenge, automated heat detection systems were developed.
- And these systems are increasingly adopted. In the Netherlands, 20% of farmers have such a system which often involves a sensor attached to a leg of neck collar that monitors a cows activity. It generates heat alarms that farmers receive after which they have to confirm the alarm using the cows’ history or look for oestrus behaviour. Just looking at the adoption rate of automated heat detection, we can say that this is a success story, but it is only a success story when it all works. And when it all works, these systems still miss some cows in oestrus and they do generate false alerts. Technical problems, computer shutdowns and simple human error reduces the success of these automated heat detection systems when applied on farm.
- How does progesterone fits in? Progesterone is a key hormone in normal oestrus cycle of a cow. Has been and is used as golden standard to define when a cow is in oestrus or to confirm pregnancy. The oestrus cycle normally takes around 21 days and at estrus, when a cow is ready to conceive, progesterone levels are at there lowest point. After oestrus, progesterone levels start to increase again. When insemination is successful, progesterone level remain at high levels. When oestrus is missed or insemination was unsuccessful, progesterone levels decrease again and this is followed by a next oestrus. Around oestrus the cow will exhibit oestrus behaviour, which is different from her normal behaviour. It is this change in behaviour that is used by most detection systems as indicator for oestrus. This picture shows the normal progesterone profile. However, sometimes these profiles are abnormal for one reason or another. We don’t know whether these abnormal profiles affect oestrus behaviour and therefore indirectly affect automated detection.
- So, what are the aims of this study.
- Second part: what did we do to get insight in the aforementioned points? Data from 2 farms located in the Northern part of the Netherlands. From these farms, selected total 31 cows, 40-70 days in milk, not yet inseminated. Milk samples 24 consecutive days. Farm A milked 450 cows conventionally and 12 cows were sampled during morning milking. Samples involved residual milk. Farm B milked 250 cows robotically. Shuttles attached to units, programmed to collect a whole milk sample from the first milking of a day from 19 cows. Farm A had two automated heat detection systems generating oestrus alerts. All 12 cows were fitted with both systems. Farm B also two systems, one was same as one on Farm A, fitted to 8 cows. The third system was fitted on 19 cows. We asked farm staff to record visual observations of cows in oestrus At the start of the study, the average DIM was 44 for Farm A and 53 for farm B. The average parity was 5.5 for farm A and 2.4 for farm B.
- All milk samples were analyzed for progesterone using a commercial on-farm progesterone kit. Analyses was done three times a week and included the milk samples that were collected the forgone 1 or 2 days. From these measurements, cow individual progesterone profiles were created . These profiles were used to determine the day of oestrus, using the manual description. This progesterone heat was considered gold standard in this study.
- So, the last part, what were the results? Based on progesterone measurements we found 30 heats from 30 cows; one cow on farm B had no progesterone heat and was excluded from further analyses. Farm staff observed 15 cows in heat, system A, B, and C generated 14, 12 and 31 oestrus alerts. Of the 30 progesterone heats, farm staff observed 3 cows as in heat on the same day. Nine progesterone heats received a oestrus alert by at least one system. To account for mismatching the true day of progesterone heat, we widened the time-window with one day. This resulted in farm staff observing 9 progesterone heats correctly. 17 progesterone heats were correctly found by at least one system. Outside that time-window farmers observed 6 more cows in heat, and system A B and C generated 4, 5 and 18 alerts. These observations and alerts were considered false positive in this study.
- To gain insight when systems generate oestrus alerts, we graphed alerts around the day of progesterone heat or day 0. System A generates most alerts at the day of progesterone heat, but it already starts with alerts two days before the day of progesterone heat and it continuous to generate alerts up to 3 days after the day of progesterone heat. Outside this period, thy systems generates no alerts.
- System B similar: it starts pinging 3 days before the day of progesterone heat and continuous to do this until 3 days after the day of progesterone heat. Outside this period, no alerts were generated. Fewer alerts were generated than System A.
- Systems C is different. Again, most alerts are generated around the day of progesterone heat. However, it also pings a lot outside that 3 day period around the day of progesterone heat. This does not involve a single cow, but the systems generates alerts for more than just one cow.
- Farm staff observed most heats around the day of progesterone heat. However, they also ‘saw’ cows in heat when they were certainly not, based on their progesterone profile. These observations may be triggered by alerts from a detection system which may bias the farmers’ view. Most alerts and observations, around 85%, happen within a 3 ay period around the day of progesterone heat. Just a quick note on this peak here. We considered this alert and observation as false, but timewise it may already be an alert warning for the next progesterone event
- So, what happens when we combine alerts from the systems using a oneday time window around the day of progesterone heat? On farm A, system A found 5 out of 12 progesterone heats whereas system B detected only 3 progesterone heats. Combining them would result in one additionally detected heat as one heat was detected by system B and not by system A. On Farm B, system B found 2 out of the 18 progesterone heats, and system C found 9 out of the 18. Mind you that system B was not fitted on all cows so this 11% does not report a fair performance. Combining the systems did not result in additionally P heats detected; all heats detected by B were detected by C and vice versa.
- Then, lastly, the effect of progesterone profiles on detection. Here you see the average profile from all 30 cows with a progesterone heat.
- Overlapping the profile of those progesterone heats that were detected correctly, we see a very similar patters....doing the same for the non-detected cows results in
- Also....a very similar progesterone profile and there appears to be no difference in progesterone profile between cows that were detected and those that were not.
- Brief capture of the main results: All 3 systems performed less than expected. Studies report a sensitivity of 80% whereas we found 25-50% sensitivity. We thought hard about reasons explaining this but couldn’t come up with an outstanding explanation. Farm staff missed about 50% of the progesterone heats that were correctly found by the systems. This result was surprising for me. The difference may be explained by farmers simply not checking all alerts or that behavioural changes already passed by the time farmers check the alerts. Most alerts and observations are 3 days around the day of progesterone heat. So, they all see ‘something’ happening in this period, but it is simply not accurate enough. Finally progesterone profiles did not differ between detected and non-detected cows.
- What I would like you to remember from this study: farmers miss correctly identified heats by detection systems progesterone profiles appear not to affect oestrus behaviour, and thus not indirectly affect performance of detection systems. Future research should confirm results from this small study. And perhaps should work with sucessful inseminations as golden standard to crank up numbers rather than use the very expensive progesterone measurements
- To finish off, I have a few acknowledgement as of course this study was a result from collaboration between many people. First of all, I’d like to thank the farmers, farm staff and two students from Van Hall –Larenstein for their hands-on contribution in collecting the data Smart Dairy Farming is acknowledged for funding this study. That was it, thank you for your attention.