The Slime Mould Collective

An international network of/for intelligent organisms

Hey slime fans,

I haven't posted up much recently but I've been busy with all things plasmodial. My dept has said they'll support me publishing what was initially a hobby; so suddenly I have work to do - especially since they let me buy a massive parallel processing laptop ( I say laptop, it weighs several kilos and will burn your legs, but it folds in half. ).

Thanks to journal pedantry this means I can't show off much stuff but there's some visually interesting stuff coming out of this and hey,we have artists and who doesn't like slime themed eye candy?

I'm now filming under amber light - the same colour as the slime - which means they don't respond to it. It also means they're barely visible. This is one square from a time laspe sequence of 25 4x4cm squares - 3000 frames in all. 

So I have work to do. As long as something's visible it means there's a signal I can extract ( that is my job, half of it at least ).

I used the first frame as a background and just subtracted it from all the other images making a difference map. The things are lit in bright orange and they're filmed on an RGB camera so the next obivious thing to do is just discard the G & B parts of the signal giving me only stuff that's 1. Red and 2. has moved during the experiment. 

Here's a sideby side pre and post processing. - the other thing that moves in this sequence is the rapidly shrinking agar - I can just crop round that. 

More processed frames - as a negative the slimes come out blue and look pleasingly like lightning bolts. 

Contour plot of a frame - useful for working out where the thresholds should be set but generally it just looks kinda neat and might be a starting point for tattoo design.

Neat trick - if you shine light through a slime ( transmitted light for those in the know ), the amount absorbed is proportional to how much pigment is present ( well duh... ), the amount of pigment is directly proportional to the thickness of the slime -

Translation - a transmitted light image of a slime mould contains 3d information that can be used to reconstruct the shape - I've exaggerated depth here because, why the hell not? It doesn't need to be ploted in orange but I like orange. 

You can do all this on a fairly basic computer with free software ( I admit I'm using Matlab which ain't free but check out Imagej ). I'm happy to show anyone how.

The only reason I need something mega is well, 25 squares per frame, 3000 time points - thats 78000 individual pictures to analyse - per run - I have 24 strains to go through at 4 strains per run ( plus Mazie as a control ) and I have to do repeats. I'm looking at in the region of 4 million frames, and that's just the first experiment. 

Views: 236

Reply to This

Replies to This Discussion

Wow! Congrats on convincing your dept. to support your work! How are you keeping the slime molds contained in the square?

That bit's easy - I threaten them with drowning. - the squares are 15mm thick and the tray has a shallow pool of water. Dries up a bit after 3 days and they do wander off then but it's the easiest way to shoot with an open container

Another run - this is all about removing everything but the slime from the image, This went blatantly wrong but it's just screaming at me to do some sort of a print

Yay! Near as dammit essence of pure slime!

Wow! How did you do that?

If you want the matlab code you can have it but basically -

Low light camera images have a shed load of noise, noise is random but signal isn't (hopefully) so I averaged the pictures over five frames ( going from one min to five on the interval)

Next subtract the background (based on an early frame with no slime movement) and separate the yellow signal.

Sometimes sharpness isn't helpful - there's still some salt and pepper noise, by blurring ever so slightly (median blur) I can cut that. That gives me a gray scale image.

Since what I'm looking at is movement the last step is to take an average frame and subtract the preceeding frame so all I see is what's moved between images.

This is no where near done - it's got the images clean, the next step is start analysing - speed, shape, preferred direction...

You could do all this in imagej (which is free) - happy to walk people through it

I should be in the lab right now but I'm not in the mood for handling lethal chemicals at the moment

RSS

© 2024   Created by Heather Barnett.   Powered by

Badges  |  Report an Issue  |  Terms of Service