Somatic Cell Count for tracking dairy cow health

[Don Spaulding]

Hi there,

I'm a total noob at spectrometry, but I understand it's one of the methods milk analysis labs use to count somatic cells in cow milk.  I had started doing research on building a spectrometer several months ago, and was sure I could handle the data side of things, but wasn't sure how to go about building the sensor itself.  I still haven't built the countertop unit yet (though the webcam is on order!).  I had some basic questions, but first, let me share some background information.

Somatic cell counts (SCC) are a count of the number of a specific type of white blood cell, which are always present in milk.  Somatic cells are produced by the animal's immune system to fight off infections.  If a cow is fighting an infection, the counts of these cells rise dramatically.  Counts are taken at various points in the milk distribution (Per cow on a monthly basis, per farm as the milk is picked up in bulk, more or less continuously by the distributor/FDA) and any dramatically high counts can require the affected milk to be dumped.

My goal is to design a sensor that can sense SCC per cow on a per-milking basis, so that anomalies in the SCC are detected quickly and animals can be treated for infections as soon as it becomes apparent they have one.  This is a solved problem for big milking companies that use extremely high priced robotic milkers, as discussed in a paper here:  http://eprints.lib.hokudai.ac.jp/dspace/bitstream/2115/33889/1/kawamura.pdf

So, my questions (some of which I know I just need to determine via experimentation) are:

Does this look like something the countertop spectrometer could handle? (sensing in the 600-1050nm range)

Would it require a special IR LED backlight, or does a typical cfl, halogen, or flourescent already give off IR light?

What's the optimal amount of liquid to shoot through?  Or: how do I know whether to try shooting through milk vs. just positioning the spectrometer over a well-lit dish of milk?

Sorry if these are simple questions to answer on my own, I'm not there yet!

Thanks.

--

Don Spaulding

Web Developer

[Jeffrey Warren]

Hi, Don - sounds like a fascinating use, and it's entirely new to me. I'll answer some of these questions but especially as we're getting lots of spectrometry questions, might I ask you to add your questions and the answers to this page on the wiki: 

http://publiclaboratory.org/wiki/spectrometer-faq

I'm thinking that as we answer initial questions we're sure to find newer questions, and it'll be good to keep a running list of them. Also it's a good way for newcomers to find an interesting question to try to answer...

Does this look like something the countertop spectrometer could handle? (sensing in the 600-1050nm range)

In terms of range, it will probably work -- we're still figuring out how to get clean data from 800-1000 nm since it can overlap with the 2nd order spectrum (there are a few mailing list conversations about this but the best I've heard is that we should be able to trust the red channel as the doubling will mostly occur in the blue channel). Whether the exposure measurements are linear is a question we still have to investigate, and a staged dilution series of samples has been suggested.

 

Would it require a special IR LED backlight, or does a typical cfl, halogen, or flourescent already give off IR light?

Halogens and xenons, and actually incandescents, all give off plenty of IR.

 

What's the optimal amount of liquid to shoot through?  Or: how do I know whether to try shooting through milk vs. just positioning the spectrometer over a well-lit dish of milk?

The easy answer is that you should try to maximize for the dynamic range of the sensor -- the reading should average around 50% exposure so you get clean data for your maxes and mins. No channel should be "blown out" (and Spectral Workbench now detects that and warns you). The hard answer is that it may be difficult to get quantitative about how much light you're actually using. Try to use little enough light that you aren't blowing out the sensor even with no sample, and little enough sample that you are still getting enough light to get a good spectrum. 

 

[Don Spaulding]

On Sat, Sep 15, 2012 at 9:33 PM, Jeffrey Warren <je...@publiclaboratory.org> wrote:

Hi, Don - sounds like a fascinating use, and it's entirely new to me. I'll answer some of these questions but especially as we're getting lots of spectrometry questions, might I ask you to add your questions and the answers to this page on the wiki: 

http://publiclaboratory.org/wiki/spectrometer-faq

I'm thinking that as we answer initial questions we're sure to find newer questions, and it'll be good to keep a running list of them. Also it's a good way for newcomers to find an interesting question to try to answer...

Sure, as soon as I get some free time, I'll see if I can find a good spot for my questions and your answers.

 

Does this look like something the countertop spectrometer could handle? (sensing in the 600-1050nm range)

In terms of range, it will probably work -- we're still figuring out how to get clean data from 800-1000 nm since it can overlap with the 2nd order spectrum (there are a few mailing list conversations about this but the best I've heard is that we should be able to trust the red channel as the doubling will mostly occur in the blue channel). Whether the exposure measurements are linear is a question we still have to investigate, and a staged dilution series of samples has been suggested.

 

Would it require a special IR LED backlight, or does a typical cfl, halogen, or flourescent already give off IR light?

Halogens and xenons, and actually incandescents, all give off plenty of IR.

Good to know, I just picked up halogen bulbs today. 

 

What's the optimal amount of liquid to shoot through?  Or: how do I know whether to try shooting through milk vs. just positioning the spectrometer over a well-lit dish of milk?

The easy answer is that you should try to maximize for the dynamic range of the sensor -- the reading should average around 50% exposure so you get clean data for your maxes and mins. No channel should be "blown out" (and Spectral Workbench now detects that and warns you). The hard answer is that it may be difficult to get quantitative about how much light you're actually using. Try to use little enough light that you aren't blowing out the sensor even with no sample, and little enough sample that you are still getting enough light to get a good spectrum. 

Thanks! 

 

-- Post to this group at publicla...@googlegroups.com. To unsubscribe, email publiclaborato...@googlegroups.com. Options at https://groups.google.com/d/forum/publiclaboratory?hl=en