Hi - I am trying to build a camera module that is capable of doing cell counts in a fluid channel. I was thinking of using an arducam or something similar attached to something like a raspberry pi for counting cells that are traveling through the fluid channel. I will be imaging at a set distance from the lens (that will not change) and a set illumination (that will not change). My assumption is that I need an autofocus capable system, but other than that, I’m not sure what my design specs should be. My thought was that I could use the 64 MP autofocus camera module (https://www.arducam.com/arducam-1-1-32- … ry-pi.html), but I’ve read elsewhere that it might be better to go with a different setup which brought me here to OpenFlexure. I’ve looked through the openflexure project specs, and it seems too complex for what I’m trying to achieve because of the moving parts, the need for multiple lenses, etc. Is there a recommended approach to distill this project down to a form factor / minimal spec project that will still give me adequate images? Thanks for your thoughts!
What magnification are you looking for? Another way of looking at it is, if you were using a conventional microscope to look at the cells, what objective would you need?
The options are different if you want the equivalent of a 20× objective (or less), or the equivalent of a 40 to 100× objective. The OpenFlexure system could do either for your setup. Both the imaging and the illumination are simpler for 20× or less.
Unless your required field of view is more than 5cm across, a standard autofocus camera module will not work. In the Upright version of the OpenFlexure Microscope, there is a separeate z-axus module that is mounted above the stage. This is what I would use in a setup like yours, on its own without the main stage.
Are you wanting to focus on small cells in a deep channel as they go by? That sounds hard. If you need to be focused in the channel, but you don’t need to change focus once it is set up, then the separate z-axis with manual operation would possibly be fine. The drift over time is usually very small, although you still probably lose focus if youvare at 100×.
Thank you William for the thoughtful response and questions! Helpful already to help me clarify.
1. Magnification around 100-300.
- Thank you for the tip on autofocus, FOV required is significantly less than 5cm in diameter.
- I don’t need to focus on a depth channel. We can assume that the depth will be ~ 1 cell layer deep.
The other challenge is that this needs to be enclosed in a housing that permits no adjustment of the stage/lens after placement. I will have to do all the optimization / design / hardware placement while the housing is open, and then once the housing is closed, no additional adjustments are possible (it will be a closed system).
When you say 100-300×, do you mean a 10-30× objective with ×10 eyepieces on a conventional microscope? So around 2 micrometre resolution and 1mm field of view?
Yes, exactly!
For that kind of resolution, the ‘low-cost’ optics version of the OpenFlexure sysyem is worth a try. It has approximately the same field of view and resolution as a conventional microscope with a 20× lens. The depth of field is big enough that you might well be able to set the focus and then just leave it. There is a thread somewhere on the forum using that optics system to look at cells in an incubator. That did not need re- focusing during 24 hour studies, and I think also not between studies either.
With the Pi Camera v2 and the OpenFlexure software you have colour correction and flat field in live view. You can do the same thing with a USB camera, but you would need to deal with the image processing yourself. There are versions for the Arducam B0196 and Logitech C270 USB cameras in the alternative optics section of Customisations and Alternatives in the Microscope instructions. The Arducam is the same sensor as the Pi Camera v2, just with a USB interface.
Thanks William. Will look into those links and hardware. I’ve been recommended to use a global shutter with a CS/C type lens ( Raspberry Pi Global Shutter Camera - PiShop.us + Waveshare 100X Industrial Microscope Lens, C/CS-Mount - 24229 ) or the HQ camera with a similar lens but would then need to adapt the software as you mentioned.
It depends on your key requirements. If the cells are moving past very quickly then global shutter might be necessary. If you have stained cells so that colour is important over the whole field of view then the HQ camera will help. Otherwise the Pi Camera v2 will perform very well.
The HQ camera has some support in the current development alpha of v3 of the OpenFlexure software. There is a prototype version of the optics module to take the HQ camera, microscope objective and tube lens.
The lens system that you have linked to seems to be quite low magnification, only 1.8x onto the sensor. The HQ camera pixels are 1.5um across, but in Bayer groups of 4 the effective pixels are more like 3um. So if the lens is optically excellent the best resolution would be a couple of micrometres. The numerical aperture is not stated, so one cannot calculate the expected resolution.
Thanks William. Color is not important. From my understanding of cell pathology, I am looking for a numeric aperture of ~0.25um to enable RBC counting. My challenge with the openflexure setup is that the housing won’t fit inside our hardware enclosure. Ideally I’d like to just use a camera module and a c/cs mount lens attached to the module without needing a separate lens or a spacer so I can fit the module/lens combo inside the enclosure, and mount the pi somewhere on the outside of the enclosure.
From some other recommendations, I have seen setups that utilize this lens:
https://www.seeedstudio.com/Microscope-Camera-300X-C-Mount-Lens-for-Raspberry-Pi-High-Quality-Camera-p-4627.html
With either the HQ, the GS, or the monochrome GS with USB: https://www.e-consystems.com/1MP-USB3-Globalshutter-Camera.asp
What gives me pause is the low MP count relative to the 300x lens on the monochrome GS options. For the HQ, it seems possible to mount the flow cell vertically to take advantage of the rolling shutter trajectory, but that will need further experimentation.