The Moos Room™

00;00;12;27 - 00;00;36;17
Emily
Welcome, everybody, to the newsroom. It is, in fact, a Christmas miracle. I am here two weeks in a row with Bradley. Which is always very exciting. And we are joined once again by a guest this week. So we have Drew Swartz here with us today. He is a grad student at the University of Minnesota working with, Gerard Cramer.

00;00;36;19 - 00;00;43;12
Emily
And so, drew, we're excited to have you here today to, to talk about all sorts of things. So welcome.

00;00;43;14 - 00;00;45;28
Drew
Yeah. Thank you for having me.

00;00;46;01 - 00;00;51;19
Brad
I'm actually excited to be here with Emily again two weeks in a row. That's like a record.

00;00;51;21 - 00;00;54;03
Emily
And again, Christmas miracle. You know.

00;00;54;03 - 00;00;56;14
Brad
That's right. Exactly.

00;00;56;16 - 00;01;15;01
Emily
All right, well, you know, drew, before we really dive into getting to know you and talking about some of the research and work you've been doing, and I know you are a listener of the Newsroom. But we do have our guests super secret questions to ask you. So we are going to start with those today.

00;01;15;02 - 00;01;33;07
Emily
So, drew, despite what Bradley will tell you, there are no wrong answers. Bradley has it in his mind that there are some right answers. But we accept all answers here. So your first question, drew, is what is your favorite breed of beef cattle?

00;01;33;08 - 00;01;50;04
Drew
Yeah, for beef cattle. I'm just going to have to keep it simple and go with Angus, as we did, during my undergrad, I would do some research with, Angus beef cattle. And it's one of my very beginning, research interests or research entries. So, yeah, I'll stick with.

00;01;50;04 - 00;01;52;19
Brad
Those. We'll accept that today.

00;01;52;21 - 00;01;54;25
Emily
Even though I didn't.

00;01;54;28 - 00;02;22;11
Brad
Exactly. Well, it seems like Angus is pulling ahead once again at 19 for the top. Herford at 14. Black barley five. Scottish Highlanders five. Charlie, four. Red Angus, four. Shorthorn, three Simmental, two Belted Galloway two, Brahma two and a whole bunch with one stabilizer. Galvin, senior in the lorry jersey, Belgian Blue Brangus, Piedmontese White Park and miniature Scottish Highlander.

00;02;22;17 - 00;02;25;13
Emily
All right, so there are the.

00;02;25;13 - 00;02;37;01
Brad
I have nothing else to say it. The Black Angus keeps pulling ahead. We need some more room for guests, if you like reference, please contact me soon to be on the podcast.

00;02;37;03 - 00;02;44;28
Emily
All right, well, now we'll move on to our second question, which is drew, what is your favorite breed of dairy cattle?

00;02;45;02 - 00;02;50;07
Drew
Yeah, I'm a big fan of Holsteins, but I think my heart has to go to the red and white Holsteins.

00;02;50;09 - 00;02;56;00
Emily
Really? White's all right. I think Brad will accept that as long as it's not black and white. Right?

00;02;56;07 - 00;03;04;20
Brad
Yes, we will accept that today. That definitely, definitely. And that is increased, red and white numbers for us.

00;03;04;27 - 00;03;06;12
Emily
Yeah, it's been a while.

00;03;06;14 - 00;03;34;27
Brad
It has been a while since we had a red and white, but black and white. Holstein still in the lead with 25 jersey. The correct answer is 20. So we're gaining Brown Swiss 13 one yard, three Dutch Belted three, Guernsey three, Ayrshire three, red and white three and Normandy two. So a wide variety people that there are few people that look for other breeds besides black and white Holstein, but.

00;03;34;29 - 00;03;35;28
Emily
A few, not many.

00;03;36;04 - 00;03;39;17
Brad
I know there are a few of those out there.

00;03;39;20 - 00;03;59;10
Emily
Yes. All right. Well, you know, now that we have the serious business out of the way, we we can move on to to our other items of business here. So again, we have drew with us, this week and, you know, Bradley, this this is a guest that you arranged. And I was really excited when I saw drew on the docket, for this week.

00;03;59;12 - 00;04;07;23
Emily
So I'll maybe let you kind of intro why you want to drew on here? And then we'll we'll finally let drew do some talking.

00;04;07;24 - 00;04;31;04
Brad
I've known drew for a while. Drew did his master's degree in animal science at the University of Minnesota with Mark Andrus. And you looked at robot feed, pushing, right, and looked at how feeding times or that helped increase production. I guess, but if that's my recollection.

00;04;31;07 - 00;04;54;07
Drew
Yeah. So it was more working with automatic milking systems and looking at the different management practices they had on the farm. So some of the things they were looking at was looking at the feed pushers and like other management practices with no view. And then we also did a feed trial within one of the milking one of the milking farms, one of the farms with automatic milking systems, where we, there are feed pellets that he.

00;04;54;09 - 00;05;19;01
Brad
Could see I was wrong. Oh. Admit that. Well, I had the gist of it, but it's been a while. You have actually. You finished a few years ago with your master's degree and are now working in the vet school at the University of Minnesota with Doctor Gerard Cramer, who does a lot of work with lameness and hoof trimming and have ventured off into a new project.

00;05;19;01 - 00;05;49;04
Brad
And that's why I thought we would have drew here today. He actually just published some research this year in the Journal of Dairy Science on his work, but it really involved looking at cameras or autonomous cameras, for detection of lameness in dairy cows and compared it to hoof trimming records. So lameness is always a hot topic. In the dairy world, no matter if you're conventional or grazing or you name it.

00;05;49;04 - 00;06;19;11
Brad
And lameness has always been an interest of mine, and I've studied a little bit of lameness from a grazing standpoint. But everybody knows me. We like sensors and technology, so why not use cameras to try and detect this? So tell us a little bit about the camera system first. So it is from cattle. I have an that is a company based out of the UK, that records cows as they come out of the milking parlor.

00;06;19;13 - 00;06;43;13
Drew
It's pretty simple, system for how it works, or at least for, like the, setup, how it goes. So the way that they have it is they have the milking parlor, and then outside of the milking parlor, above the ground looking down onto the cow. So it's going to observe the cows walk under the camera. So there's a camera above the cows on the return alley, and then the cows will walk about 3 or 4m under underneath the camera after milking.

00;06;43;15 - 00;07;00;13
Drew
And it'll record a video. And then it'll send the video off to, wherever they keep their software. So it's more of a cloud based system that they have going on. And then the their AI algorithm will look at the footage and then assign a mobility score. So the way that they do it is from 0 to 100.

00;07;00;15 - 00;07;21;13
Drew
But they also have a way of doing body condition scoring as well. So they also produce body condition scores. And then the way that it's recognized is through, the RFID collar that the cows wearing. And then along with, recognizing the patterning of the cows. So it pairs them both at the same time. But then how it the specifics on that, on that tissue.

00;07;21;16 - 00;07;27;11
Drew
But it does a little bit of pattern recognition and then, using the RFID to let them know they're there.

00;07;27;17 - 00;07;50;17
Brad
Yeah. That's interesting how it, it records all of that information and obviously, like most technologies, they're all proprietary algorithms. And nobody wants to tell you how they figure it all out. But it is able to detect lameness or at least try to detect lameness based on the gate of the cows. So then how many farms did you work with on on your study?

00;07;50;20 - 00;07;58;14
Brad
There was three farms. Correct. And were they all here in the Midwest or were they across the US?

00;07;58;16 - 00;08;01;02
Drew
Yeah. There were three farms in the I was in the Midwest area.

00;08;01;02 - 00;08;06;23
Brad
And they were all Holstein. Right. They're all Holsteins. Nothing against the Holsteins.

00;08;06;23 - 00;08;19;21
Drew
But I will say that we, only pulled it from the Holsteins because everything's been worked around looking at Holstein. So all the validation techniques were looking at Holsteins. So we decided to keep it within the whole thing.

00;08;19;24 - 00;08;41;12
Brad
And I think there's there's ulterior motives to all of this with working with the US Council on Dairy Cattle Breeding to develop a genetic evaluation for lameness. But we can we can discuss that at the end for a little bit. So you worked with hoof trimmers and to capture their data based on what they recorded at, trimming time.

00;08;41;12 - 00;08;42;00
Brad
Is that correct?

00;08;42;00 - 00;08;59;09
Drew
Yeah. So it's kind of more of, just we ended up going through like their dairy management software as a dairy com for most of these farmers. And what we did was we ended up pulling their host trimming records. So seeing what they put into the, dairy comp systems. And then we took that as like day zero.

00;08;59;09 - 00;09;16;01
Drew
And then we looked at how the mobility scores leading up to that event were, and kind of just compared them between whether or not cows had lesions and didn't have lesions and then also like kind of dug deeper in those like to see how the trends leading up to the actual hoof trimming events were.

00;09;16;03 - 00;09;25;18
Brad
So that so the system will give a score every, every time the cow walks through a the camera system. Right.

00;09;26;09 - 00;09;28;12
Brad
So then it has the opportunity to.

00;09;28;15 - 00;09;45;24
Drew
Okay, okay. So for some cows we and that we ended up doing it at a weekly value. So we ended up doing like we average the scores for the week and took them from a weekly average because some of the cows are missing data for daily data. So I think the intention is they go under the camera and they should be recognized, but sometimes they might not.

00;09;45;29 - 00;09;56;29
Brad
Right? Yeah, there's a lot of reasons why they might not get recognized. Maybe they lost their RFID or they went through it too fast or or you name it. I think there's a lot of a lot of those issues.

00;09;57;01 - 00;10;26;06
Emily
So this leads into one of my questions. And it's it's kind of a more basic question. But yeah, could you just maybe explain, like how how do we actually get this data? Like how do we get the cows through the camera? Is it put in a holding pattern or something like, is it a thing where, yes, we make sure every cow goes through it every single day if we can, or, you know, kind of the, the mechanics of actually getting the score right and getting that scan done.

00;10;26;13 - 00;10;44;27
Drew
Yeah. So I don't think it's anything out of the ordinary. It's going to be mainly for the lactating cows, as it's the cows that are leaving the milking parlor. So as long as there's a return alley that the cows are guaranteed to walk through or walk down, all the cows walking out of the parlor will go under those RF IDs for under the camera.

00;10;45;04 - 00;10;58;06
Drew
And then we have a farm has to return alley. So they have two cameras. And then after that you can set a timing for it. So you can either have like continuous recording if you wanted to do like two times milking per day, three times milking per day. Or you can just set it up for like a shift.

00;10;58;06 - 00;11;17;04
Drew
So some of the farms we do is just like they have to go for eight hours, they do three times a day milking kind of thing. So then after that the algorithm or the once the milking is done, it'll produce a score by the next day. And then you can just log on to their website or like their platform, or you can have it incorporate it into their comp and you can get the scores like that.

00;11;17;09 - 00;11;38;24
Brad
So it will give you a score every day. I didn't realize that it would integrate back into dairy comp, so farmers could actually look right at dairy comp and figure that out, which makes it a lot easier to brand is not a big fan of having multiple systems and computers in barn offices. I have four computers in our barn office that are running different technology and it is maddening.

00;11;38;29 - 00;12;07;04
Brad
But anyways, that that's a different story for another day. You looked at all these trimming records. Do you know what? So what is the biggest issue, on farms that you worked with? Is it digital dermatitis or efforts, or is it solar ulcers or white line disease? What what were the biggest lesions on on these farms? I'm just curious to to to remember what that might be.

00;12;07;10 - 00;12;33;12
Drew
There's kind of variation between the farms, but the ones that we see the most, in the study was digital dermatitis. And then so that's kind of the more of the infectious of lesion. And then we also looked at white line. And so ulcers which were the like the noninfectious lesions. But then one of like the challenges in the study is that when we were like looking at things in the past that since we took five years of data, we had to calibrate them, we had to calibrate the hoof trimmers.

00;12;33;12 - 00;12;44;24
Drew
So kind of had different meanings in the beginning and things like that. I mean, you know, that analysis a little hard, but but yeah, I'd say the answer your question. Dermatitis was the most common lesion that we had.

00;12;44;24 - 00;12;53;24
Brad
So how did you get to calibrate hoof trimmers. We're sometimes they didn't call some things lesions or not or how how did that work out.

00;12;53;27 - 00;13;14;12
Drew
Yeah. So I would have to say this part was less on me and more on Girard. So before I got to the lab, you was doing this calibration part since it was five years of data and I wasn't here yet, but I think it was more making sure that we called everything the same thing. So like with digital dermatitis, you can call it digital SD, strawberry foot and the name goes on.

00;13;14;12 - 00;13;30;26
Drew
So when it comes down to like recording it in the software is that they're using you can have different names. So that was just more difficult on the back end. And then also with the different softwares that the trimmers were using, just kind of getting everybody on the same page or like how we should call it and like the locations.

00;13;30;26 - 00;13;37;12
Drew
So basically look at like taking the perception from the, the iCar, at least for home health and going out for that.

00;13;37;14 - 00;14;05;11
Brad
Okay. So let's get into the nitty gritty. And what did we all find. So you tried to figure out whether the cattle I was trying to predict lesions or some sort of lameness in these cows, did it do that or how well did the system work at at recognizing either locomotion changes or any sort of trends with hoof lesions?

00;14;05;12 - 00;14;23;01
Drew
This is where it gets tricky. So we look at things like whether or not the cow that had the hoof churning event, let's say, whether she was a cow that went to the hoof trauma and she had a lesion or she did not have a lesion and she just received the hoof trimming. So we'll call those cows the trim cows and the lesion cows.

00;14;23;04 - 00;14;42;29
Drew
So just like at the broad level between them, when we kind of looked at the scores, what we saw is that like the at the population level, I think this is very important. At the population level, we saw that the median score was kind of consistent throughout all four weeks leading up to the hoof trimming event. But then for the, cows that had hoof lesions, we saw that the median score increased.

00;14;42;29 - 00;15;02;15
Drew
So like their distribution was shifting upwards. But I think the challenge here is that when it comes to like using this technology and like looking at an individual cow level instead of a population based level, it's like where do you draw that line for the change in score to get the cow to go to the whole stream? Or should we use a specific cut?

00;15;02;17 - 00;15;29;02
Drew
If we use a specific like numbers like, say, 5050 or greater than the count of the whole tumor, we have a lot of false positives going to the customer and right, that costs money. And we kind of want to figure out a way how to make it more efficient. But so I would say that we did see for the cows that did not have lesions in the cow that had hopefully lesions, we did see a pattern that cows that we're going to have that had lesions that the host from event had, like their distribution of scores increase.

00;15;29;04 - 00;15;43;13
Drew
But then it also gets challenging as what we saw about when we looked at the specific types of hoof lesions, they all had different patterns. So it's going it's kind of figuring out how we can use these scores to get at detecting these lesions.

00;15;43;17 - 00;16;10;29
Brad
I think that's interesting that you see different patterns with the cameras with different lesions. And I would probably guess that some of the ins, especially where they're located in the hoof, might produce a different gait than some other one. So you get different patterns and different scores based on that. So I can see where it might be difficult to try and figure out what what lesion is happening or how lame those cows are and what the cut point is.

00;16;11;02 - 00;16;36;05
Brad
I think we see that in all technology where you might be getting some false positives and you don't want to bring, obviously bring cows to hoof tremor that doesn't really need to and will cost extra money. But it is always interesting when I think about when I talk to farmers about precision technology, especially with activity monitoring systems, trying to detect disease.

00;16;36;08 - 00;16;57;11
Brad
They also want to know how many days before can I detect a disease, how many days? And was there any not to put you on the spot? But what what are the number of days before something happens? Or could you be able to tell that? Or were you seeing trends? Maybe, maybe a week before this happened? Or did they come on all of a sudden?

00;16;57;13 - 00;17;03;03
Brad
I think every cow is different, but was there any trends that you'd notice there?

00;17;03;05 - 00;17;23;10
Drew
Yeah. So when trying to look at like whether or not how early the cows or the how early producers can start detecting animals, I would say that we don't really have the data on that one yet. To answer, because I know we're in trying different algorithms or like trying different methods, like different changes in scores and just trying to figure out how we can get like the most accurate readings.

00;17;23;10 - 00;17;26;29
Drew
But I don't think we have the ability to do that yet.

00;17;27;01 - 00;17;40;17
Brad
Where does this go in the future? What what are some things that you're going to look at with this information or this cattle AI project, and where where might farmers use that.

00;17;40;19 - 00;18;01;02
Drew
Yeah. So I think the where this can most or where going forward where I think this would be like most beneficial in like an applied perspective is figuring out how we can use the data produced by the system and then pairing it with other data collected on the farm. So we currently just ran a, a study trying to answer this question.

00;18;01;04 - 00;18;19;25
Drew
So what we had what we've done so far is we enrolled about 500 cows in a study where we host trained them or we took a sliver of their hooves and hoof tested them, make sure they didn't have any lesions. And we enrolled them into the study. So we rolled about 500 healthy cows and about 100, 120 days later, we reevaluated all the cows.

00;18;19;27 - 00;18;37;28
Drew
So then we gave these cows like a full trim, like therapeutic recorded all the hemorrhages that they've had or so at the baseline or at the first enrollment. We looked at like what was going on with the hoof, like, was there any damage, like any hemorrhaging going on? And and then if they had like severe lesions. So it's just like white line.

00;18;37;28 - 00;19;02;23
Drew
So ulcers, so ulcers, digital dermatitis. We didn't continue with those. So we wanted like cows that weren't having active lesions per se. And then we followed these cows up 100, 120 days later. And then we, evaluated their hubs again. And what we did was to see what lesions they've had. So then from so now we have 120 days of catalyzed scores roughly for each of these cows.

00;19;02;26 - 00;19;27;26
Drew
And then we also had access to the producers, dairy comp so that we pulled a whole bunch of milk data and just activity data and changes like that. So what we're working on now, and now that we have all the data and it comes the fun playground part in trying to detect lesions is now we're building machine learning algorithms to try to detect, to see if we can make an algorithm that can, like, classify cows as having a lesion or not.

00;19;27;28 - 00;19;31;27
Drew
So then we can have a applicable use for the technology on the farms.

00;19;31;27 - 00;20;02;18
Brad
Well that's good. I think there's a lot of promise for this in the future. Obviously a lot of people have tried it with cameras are probably the number one hot topic with trying to figure things out on a dairy farm nowadays, especially with feeding and lameness detection and you name it. As a follow up to this project. Actually, I know, your advisor Gerard, is working with the Council on Dairy Cattle Breeding, which does the genetic evaluations here in the US.

00;20;02;18 - 00;20;32;15
Brad
So they're trying to look at, developing a genetic evaluation for lameness scores in the US for cows and bulls based on cattle data and trying to figure that out. I know there's other countries, namely Sweden has had hoof health, evaluations for many, many years. And other countries in Europe as well. So the US is a little behind, but it's always good to see trying to develop new information.

00;20;32;15 - 00;20;39;25
Brad
So I'm glad that your research will help develop genetic evaluations for cows and bulls in the future.

00;20;39;25 - 00;21;06;00
Emily
So I have one, final question. You and Brad, we're kind of going back and forth so quickly. You know, we we like to talk about things as, as being management tools and, and the like on this podcast. And, you know, also just thinking about what is the on farm application with it. And so I'm curious, you know, what sort of feedback, if any, you got.

00;21;06;02 - 00;21;24;16
Emily
From the farmer cooperators you were working with in the study, you know, did they think it was a useful tool? You know, did they say, hey, this is something I want to have and and also, you know, what did the hoof trimmers think about it? You know, I know we have some listeners who are hoof trimmers. So I'm sure they're curious.

00;21;24;16 - 00;21;31;16
Emily
You know, what did their colleagues think about this technology as well. So so what was the feedback from the farmers and the trimmers?

00;21;31;16 - 00;21;53;28
Drew
So from the hoof trimmer perspective, I haven't really heard too much back from them about the technology itself. But I do know that the producers that we work with are still using the technology, but at the end, like we're all still trying to figure out how we can actually use this. So sure, we can get a mobility score, but then the question is, how can we use this mobility score to actually like tackle lameness?

00;21;54;01 - 00;22;14;19
Drew
Because if we're sending if we're seeing cows with, threshold above 50, and let's say that's what we're going to call lame cows of the software itself, and we're sending counsel a hoof trimmer, and they're not lame. Now we're getting charge from a health trimmer for not using or for where we're sending a cow to the hoof trimming to get trained.

00;22;14;19 - 00;22;33;27
Drew
And that's going to be an expense. And then there's also some research out there that shows sending cows of the host trimmer that don't have lesions, actually impacts like their milk production and like their behaviors moving forward. So now it's like kind of a you're not only losing money directly by paying with those trimmer, but now you have like indirect losses by not producing more milk and like messing with the cow.

00;22;34;02 - 00;22;56;09
Drew
So I think I think the main question is how can we actually use this technology moving forward to tackling this? I will say I, I think if trimmers get more mad about treating cows and then like not mad from like a perspective of like that, they're treating the cows but like it slows them down. So they're getting through less cows.

00;22;56;09 - 00;22;59;03
Drew
So host trimmers would rather trim healthy cows.

00;22;59;06 - 00;23;18;12
Brad
I agree with you there. And I've had lots of conversations with hoof trimmers about that, where sometimes it can be demoralizing to go to a farm and you're just working with cows with bad hooves all day and and a nice to have. You can trim a hundred cows a day or more, and they're all just normal. And you.

00;23;18;15 - 00;23;22;04
Brad
It's a good day to trim cows that don't have any problems.

00;23;22;06 - 00;23;41;27
Drew
I would agree. I think customers take pride in their work and like the whole health of dairy farms, I think they're a big collaborator there and I think I'm not sure if they're undervalued or not. But I know that, like I spent some time with trimmers during my PhD and like, they kind of worked with me on trimming and it's just like one of those things, like they take pride in what they do and they.

00;23;42;01 - 00;23;43;15
Drew
Yeah. So I got my real.

00;23;43;17 - 00;24;05;25
Brad
No, I agree. And I sometimes they are undervalued on farms because they're just in there to fix problems. And I could go on days about hoof trimming and and fixing problems. And that was one of the reasons why I started off trimming our dairy twice a year. Because we were having so many problems. And it just it was like getting the host trim in there to fix problems.

00;24;05;25 - 00;24;29;02
Brad
And now we can trim 300 cows in one day with a bunch of trimmers. And we have ten, ten problem cows. So it's much better now doing a lot better job on on hoof trimming. That's another subject for another day. And I got a progressive dairy article about it and got lots of good comments. So trim your cows more than just problem cows.

00;24;29;02 - 00;24;30;05
Brad
I'll leave it at that.

00;24;30;08 - 00;24;52;03
Emily
We can wrap it there. Thank you again so much, drew, for for coming on today and talking cattle, iron and hoof trimming. And I know I learned a lot and and it's always interesting to learn about the latest, technology and sensor that we're seeing in dairy farming. So thank you again so much. Drew, for stopping by the podcast.

00;24;52;03 - 00;25;19;09
Emily
If you have any questions, comments or scathing rebuttals about today's episode, you can email those to the moms room at Umkc. Edu. You can also call and leave us a voicemail of (612) 624-3610. Find us on the web extension, um.edu. That's a wrap by.