Using OFM as a Simple Automated Cell Counter

[engpol]

Simple Cell Counter for Performing Cell Dilution Calculations

Hi Everyone!

Just wanted to spread the word about something I have been working on which I hope may be of some use to people in the future.

Some quick background:

Me and a lot of people in our lab do a lot of work with a large number of immortalised cell lines (HEK-293/INS/CHO etc.). As you can imagine, we routinely have to perform cell counts of our cell suspensions when seeding into a wide range of our multi-well plates for various different experiments. For this, our PI invested in an Invitrogen Cell Countess - which I must say is really great - however it is incredibly expensive, we now have too many people performing experiments and only 1 machine meaning the demand has now outgrown supply.

Seeing that I had an OFM I built for an outreach event last year, I thought I can’t be the only one in need of something like this. Someone must have developed a build of the OFM for use as a cell counter. I had a look around the forums, and while I found a lot of cool projects doing arguably much more sophistacted things beyond my use-case (a lot of work looking at image classification etc.), I couldn’t really find anything which did exactly what I needed/I could get working on my own machine quickly/very easily.

So I thought I would make one myself.

Cell Counter

In the spirits of doing things as frugally as possible, it only requires the Basic Optics configuration, 2 relatively small extra parts to be printed, and any cheap generic 22x22mm glass coverslips to mount your cell suspensions.

Most importantly, I have included an image of my RPi with all of my code which runs natively on the RPi, meaning if you have an OFM, getting this to work in your lab can be as simple as flashing it onto your SD card, and printing 2 extra parts in PLA for the counting chamber:

Segmentation is performed either natively on the RPi (using classical computer vision on local intensity variation; ffc → scharr operator → blur → hole filling → watershed etc.), or optionally on a seperate server running cellpose (the detailed setup for which is also covered, hopefully, in detail on my github).

The usage and GUI is very user friendly, you simply:

1. Mount 50ul of your cell suspension in the counting chamber
2. Place in on the scope
3. Boot up one of the counting scripts using a shortcut on the desktop
4. Wait for the micrsocope to acquire images and segment cells
5. Use on the on-screen calculator to perform your desired diltution calculations

There are obvious limitations to the method I use to both mount and calculate the initial cell concentration (which I can discuss in more detail), however in my limited experiments, I found it works just fine/about as well as manually counting cells for cell suspensions in the ranges that you generally get when splitting semi-fully confluent T25 → T175s. Accuracy drops quite a lot below this, but this is easily solvable by acquiring a larger number of FOVs (again, as my use-case is for cell dilution calculations, I am generally counting relatively dense suspensions and speed was of greater importance hence this is fine for me)

Github

More info along with the stl files, tutorials, and RPi image can be found on my github repo, which I will be expanding on soon.

Hoping this will be useful to some people who may come across this in the future!

[Yashka]

Hey! This is very cool. How are you making sure you have a uniform chamber thickness for the suspension to spread out in?

have you tried it with a classic hemacytometer?

[engpol]

Hi, thanks for the kind comment!

So I thought about using a simple haemocytometer as the mounting chamber instead of glass coverslips, however my main problem was I was unsure if I could find a chamber that would:

1. Be affordable enough to suggest use generally (some of these can reach into 100s of £, and I was trying to keep this as cheap as possible - fwiw, when I built my first microscope, the RPi was about half the price of what it is now ).
2. Be widely available: you can find some second hand chambers cheap from Amazon/e-Bay, however these differ slightly in their exact exterior dimensions, and thus perhaps would not fit very snuggly into the mounting chamber I designed. 22x22mm glass coverslips should be widely available and exact in their dimensions.

I actually got the counting chamber itself from the mini-hematology-lab project and designed the mount around it. I experimented with a range of pipetted volumes to ensure complete coverage of the 22mm² area without any floating/loss of suspension around the edges of the coverslip. In my hands, I found 50µL worked very well in this regard, and almost perfectly covered the area of my coverslip.

Secondly, I think technique comes into play a little bit here. If you use good standard practice for mounting coverslips (place suspension in centre of plate, and then slowly place coverslip at 45o angle, wait a little while for cells to settle). You can get a pretty good/even suspension by my books. It’s not perfect, and you may get slight variation across the length of coverslip in the direction in which you placed the top coverslip, however at least in my tests, this didn’t affect the accuracy of the count all that much. As with anything, I think you can also use your best judgement a bit; If you have a relatively bright light and place the removable slide chamber after mounting on a dark book/something for contrast, you can see even by naked eye how even your suspension is. If it’s bad just try again, only takes a few seconds.

The obvious caveat here, is that I am recommending this be used only for counting relatively dense suspensions, spanning only about 2 orders of magnitude, in which the small variations in liquid layer thickness don’t contribute to the overall error (in my opinion) as substantially.

I actually think you could improve the count more substantially by increasing the number of images the microscope takes. The OFM has, from the POV of a cell-counter at least, got quite a small eFOV, coupled with a relatively slow move speed of the motorized stage. For dense suspensions I found taking fewer images is ok, as you get more and more dilute, the cell number in the 4 FOV I acquite drops down to just a 1-10 cells, where the local variation in chamber thickness, and debris mis-classification (which is why my error tends to always over-count) can baloon the error up. Like I said in my original post though, for me speed was much more important, as I don’t intend to use the cell counter for counting cells anywhere near the 10^5-5.5 cells/mL range.

However, if anyone has a source for an inexpensive, widely available, hemacytometer of standard dimensions size, please do let me know. It doesn’t even need to be laser etched really. I’ll happily order one, and design a different slide holder for it, and see how it compares. All you’d need to change in the code is the initial suspension volume parameter which would take a couple seconds.