Hello, I am new to open flexure. Interested in using a TensorFlow-Lite object detection model to count microbes. I recently developed a demo model for counting spirochete bacteria (https://spirocount.github.io/). Currently building an OFM and hoping to get dark field microscopy good enough to continue. After that, looking to build an easy to use (touch screen?) interface to display and control automated counting. I am still exploring options and I am happy this exists! Also, would be happy to collaborate!
great, welcome! Do stay in touch - dark field ought to be OK at least for low NA (e.g. 40x, 0.65NA and @samuelmcdermott has got this working with the LED array on the delta stage (it should work just the same for the regular microscope too).
I am interested too, Not sure what i can do to help out here.
1 Like
Interested and curious to what has already been done? Most interested in sticking open flexure 7.0 into an incubator with a cell culture flask or Petri dish and monitor the culture automatically using computer vision.
1 Like
I know there are a few folk who have done this (microscopes in incubators), though I’m not sure to what extent their workflow uses automated image analysis. I’d have thought the analysis may be quite standard, though, and it would be lovely to see some links between here and image.sc if you find anything relevant there. This reminds me of some of the work @quigly has done, so I’ll tag him here just in case there’s a useful connection.
1 Like
Yeah for the past few years I’ve had some scopes in my incubator to make movies of macrophages infected with yeast in 35 mm petri dishes, I have it take an image once every minute for 24 hours. I can fit four scopes easily in one incubator and could probably even get four more on a second shelf if I really tried. Overall it works incredibly well, the main thing to be careful of is that the PLA will become brittle so just be careful if you have to manually adjust any of the pieces. I’ve wanted to try making one out of other materials like ABS but still haven’t gotten around to it.
To remote desktop I just use VNC and set up a hotspot on my desktop in the next room. The IT department has set up some restriction making it so the pi’s can’t connect to the wifi so if you run into an issue like that hotspotting works great, my hotspot is in another room across the hall from the incubator and the signal is still great. Controlling the motorized stages worked fine (I used the arduino nano rather than the sangaboard version) and ideally I’ll to move up to scanning fields of view but right now I pretty much just do one static field of view and don’t have to change focus much, if at all, between plates.
I haven’t gotten very far with automating analysis though as I’m trying to track these small yeasts as they travel between macrophages and being able to segment and track all these fast moving, changing, and overlapping macrophages has been rough. Depending on what you’re doing it might be very easy, like if you’re just looking to monitor for confluence or something I think that would work well off the bat.
2 Likes
Hello, Quigly: This is a pretty good use for the OFM, in my opinion. I am a pathologist with experience with tissue culture techniques and thought to give it a shot as well. Of course, this will need some Python code which I am not really good at. In any case, I wanted to share a couple of my thoughts with you.
I would print the scopes with high-temperature PLA that can withstand temperatures of up to 80 degrees Celsius. ABS may be overly stiff and interfere with the OFM’s restricted range of motion.
Tissue cultures are useful because cells grow on a monolayer and you don’t have to focus too hard. If you have any organoid structures or cell groupings, you may “z-stack” the photos. This can be accomplished using ImageJ (fiji).
1 Like