Automated cell counting of yeast using ImageJ

Cell count automation isn’t as easy as I expected, and I’ve learned quite a bit in the processes getting it to work well.  Hopefully what I’ve figured out will help you to be automatically counting yeast cells and viability in less time than it took me!

Sample Preparation

It takes more than fiddling with the algorithm settings in ImageJ to get good results. It all starts with sample preparation.  The cells should not be clumped up and should be sparse across the field of view.  If you aren’t staining for viability 5% acetic acid (white vinegar) can be used to unclump the cells.  Unfortunately most viability stains, such as Methylene Blue and Trypan Blue are very sensitive to pH and will not be absorbed by the cells at a low pH.  Alternatively malt extract can be used.  In the presence of fermentable sugars yeast will naturally separate in order to create more surface area to metabolize the sugar.  For a good image of the yeast the cells should also be sparse enough to make counting easy.  50-200 cells in the field of view is a good target.  This corresponds to a cell density of about 10 to 50 Billion cells per liter which will appear slightly turbid in a test tube.

If the foreground stands out well from the background ImageJ can better differentiate the cells.  Contrast can be enhanced by staining.  Some common stains are as follows:

• Methylene Blue - Stains dead cells blue.  A one to one ratio of cell suspension and a 0.1% solution is commonly used.
• Iodine - Stains cellular glycogen brown.  The intensity of the color is proportional to the glycogen content.  0.2% solution is commonly used. [1] 

Imaging

It’s possible to take a picture through a microscope by holding a camera at just the right spot over the eyepiece, but it’s very hard to take a good image this way.  The simplest way to get a good picture is by purchasing a USB microscope camera.  An alternative is to use a camera adapter for your cell phone.  Most cell phones have pretty nice cameras.  The iPhone camera is certainly better than most web cameras, and even better than most USB microscope cameras.  If you have it, why not use it?  To solve this problem I used a 3D printer to make one.  It attaches to the eyepiece and uses rubber bands to attach to the phone.  

Brightfield imaging works well if the microscope is setup well.  Darkfield provides a more crisp image, but color is difficult to distinguish.  A dark field accessory is available for many microscopes.  For mine I printed a disk that can be placed in the light path normally used for color filters. 

Some tips for getting a good image:

1. Use a fixed focal length on your camera of about 10”
2. Zoom in on the camera as far as possible then adjust focus on the microscope
3. Adjust focus on the camera if needed.
4. Zoom out on the camera so that most of the field is in view.
5. Turn the brightness setting on the camera down (I use -1) to compensate for the black corners of the image.
6. Use a low iso and long exposure time for the best image quality.
7. Adjust the iris aperture to accentuate the cell membrane, but not so much as to entirely loose the cell color.  (Mine is normally almost closed)
8. Adjust the focus so that the centers of the cells are bright, but not so bright that the color is washed out.
9. Take a picture of an empty field to be used for the background removal.

Algorithm

It can take hours and hours to really hone a process, and can be quite frustrating if you are working with poor images.  I’ve played with this for way more hours that I should have, and come up with the following procedure:

1. Make a background image by blurring a copy of the specimen image.
2. Subtract the background with the image
3. Connect the shapes with maximum, minimum and blur functions.
4. Make a mask by using the Auto Threshold function.
5. Find the particles and add them to the ROI manager
6. Make the measurements.

Place these files in your ImageJ macro directory to use them.  The "Batch Yeast" macro processes all of the images in a folder with the "One Yeast" Macro.  The results are saved in a coma separated log file as well as annotated images.  Cells detected as being alive are circled in white, dead cells are circled in blue, and trub is boxed in red.

 

One Yeast 

Batch Yeast

The FIJI build of ImageJ can be downloaded here:
http://fiji.sc/Downloads

[1] V.E. Chester, Heritable Glycogen-storage Deficiency in Yeast and its Induction by Ultra-violet Light. 1967 J. gen Microbiol (1968), 51, 49-56

Starter Calculator

How time effects cell growth is a factor that I have felt is missing in the existing calculators, which led me to create this starer calculator.   Another personal gripe I have with the commonly used calculators is their unrealistic implied accuracy.  For honesty I have included one sigma error bars on the cell calculations.  I've collected data from dozens of experiments and factored it all down into this Excel calculator.

This calculator also lets you select which strain of yeast you are using.  Right now it is limited to US-05 and S-04, but I plan to expand this as resources allow.

You will need to enable macros to use this calculator.

This is the Beta Release, so give it a try and let me know what you think!

Download the Starter Calculator

Cell Density Meter Update

When the Cell Density Meter is released we want to make sure we have worked out all the bugs, smoothed all the wrinkles, and that it works seamlessly.  Our goal is straight forward operation with complex algorithms underneath to provide accurate measurements in the strenuous brewing environment.

At this point we have several working algorithms and are evaluating how they can best be implemented.  The enclosure is in its final stages of fit checks and adjustments.  

To put the Cell Density Meter though it's paces I conducted a comparative study of starter growth with and without a stir plate.  In summary the yield is about the same, but the cell growth is much faster with a stir plate.  Download the full report below.
Agitation Effects on Growth Rate and Yield of Brewer’s Yeast

Fermentation Model

For my book I developed an excel based model for yeast growth.  This was based on experimental data, physical and biological properties, and also a few "scale factors" to align the model with measured data.  Most of the data is from beers with an OG from 9°P to 16°P (1.036 to 1.064) at temperatures near 70°F (25°C) using standard pitch rates of US-05 yeast and Briess Malt extract.  The model should hold up fairly well near those operational points, but I am sure there will be some deviation in nearly every case and significant deviation in a few cases.

Description of Graph
The lines are generated by the model.  The squares and diamonds were measured experimentally.  The cells that have an orange background are intended to be changed. 

Total Sugar - Sugar concentration in degrees Plato (multiply by 4 to get brewers points)
Single Chain - Component of Total sugar that is simple sugar
Two Chain - Component of Total sugar that is more difficult to ferment 2 chain sugar.
Sugar - Measured sugar (Refractometer corrected for alcohol)
Yeast Suspended - Billions of cells per liter (as counted on a hemocytometer)
Total Yeast - Calculated total yeast population
Yeast Suspended - Calculated yeast remaining in suspension (note that this is highly strain dependent, but can be adjusted in the model)
Growth - Billions of cells grown per day.

My book goes into more detail about the application of this model and how this graph can be used to better understand yeast growth.

Yeast Growth Fermentation Model Download