Slashdot Mirror
Activity of Whole Fish Brains Mapped Second To Second
ananyo writes "Researchers have imaged an entire vertebrate brain at the level of single neurons for the first time. A team of scientists based at the Janelia Farm Research Campus in Ashburn, Virginia, were able to record activity across the whole brain of a fish embryo almost every second, detecting 80% of its 100,000 neurons. The work is a first step towards mapping the activity of a whole human brain — which contains about 85,000 times more neurons than the zebrafish brain. The imaging system relies on a genetically engineered zebrafish (Danio rerio). The fish's neurons make a protein that fluoresces in response to fluctuations in the concentration of calcium ions, which occur when nerve cells fire. A microscope sends sheets of light rather than a conventional beam through the fish's brain, and a detector captures the signals like a viewer watching a cinema screen. The system records activity from the full brain every 1.3 seconds."
Well, this is a first.
"Oh look, a hook!"
.
Each hour-long experiment generated 1 terabyte of data and they were able to detect 80% of the 100k neurons in the fish's brain. So that works out to 1 terabyte $\div$ 1 hour * (3600 seconds/hour) / (1.3 seconds / data item) / (80000 neurons) = 4513 bytes per neuron in the dataset.
Run that as
perl -e "print 1e12/(2769.23077)/(.8*1e5)"; echo 4513.88888763503I wonder how much faster the ata really needs to be captured in order to get as much resolution as needed to understand what's going on.
This is like taking slices of 80% of a computer's memory once a second. Sure, you might be able to get an idea of what's going on, but until you can see the whole picture, a lot of things are not going to make sense...
#fuckbeta #iamslashdot #dicemustdie
The next steps are fairly obvious..
1) Figure out how to write data to said mapped brain.
2) Attach USB Interface to organic storage unit.
3) Profit!!!
Since these techniques rely on bio luminescence, can the light generated from the neural activity travel to and stimulate the retinal cells? Can the animal... see itself thinking?
http://www.visual6502.org/JSSim/ in action.
They sentenced me to twenty years of boredom
Not apparent from the (cool) video they linked to is that according to the paper in Nature (yes I RTFA and I followed a link) they say they achieved cellular resolution (the video must be a down-sampled version). This would explain them collecting 1TB of data for each 1 hour "run". Another neat thing to notice is that their data is 3D, they are collecting volumetric data (as you can see from the video "slice") and explained in a previous paper. Impressive! Now if only they could increase the temporal resolution (multiple parallel scanning beams?) we could really see how a fish thinks!
They say they could collect data from (currently small) sections of mammalian brains but it would require surgery. I wonder how soon until we see monkeys with their skulls replaced with transparent plastic or glass? Maybe they could use (a very advanced version of this) on patients undergoing brain surgery.
By the way, are there any transparent plastics that are suitable for 3D printing? Biocompatible? I can see a time when some really crazy performance artist replaces his/her skull with a transparent one. I guess they would have to wear a hat whenever they went out into the sun though.
Nice work. I look forward to see the 1 millisecond time reolution. The researchers state that the human brain contain 85000*100000 -> 8.5 billion neurons. Most textbooks says the human brain has about 100 billion neurons. There are also papers out telling that the neocortex of a young male contain about 22.8 billion neurons (Pakkenberg). So the human brain is much more complex than stated.
Castle, Rock, Castle, Rock, Cat Whoaa!, Castle, Rock, Cat Whoaaa! ...
"If it's lost, it'll turn up. Things always do" "I love it when a plan comes together"
DARPA will throw money at this - if details can be refined, and RECORDED, then IMPOSING the recorded patterns onto another brain equals INSTANT CONDITIONING !
redneck geek
This is obviously the first step to being able to upload. First it will be fish (e.g. lobsters), then kittens, and sooner or later, humans. But we should make sure we get the ethics and legal aspects sorted out first, I wouldn't want to die, and then wake up a slave to someone else.
A bright new trans and/or post human future awaits us!
HELP MY ACCOUNT HAS BEEN HACKED BY AN ILLIBERAL ART STUDENT SET TO DESTROY THE INTERWEBZ!
I'm sure recording fish brain is awesome, but I can't comment on it. I've forgotten what the news blurb said.
It's not just photos (2-d) but three-dimensional volumetric acquisition, so if it's got the same resolution in all three dimensions, then each three-dimensional slab acquired every 1.3 seconds is 361MB ( = 1.2 TB / 2769.23 slabs acquired per hour) [ which is also = 1.2 TB / (3600 seconds / 1.3 seconds per slab acquisition) ]. /arithmetic or /arithmegeek, to use slashdot-speak) ;>)
.
Then, 361 MB per slab / 80Kneurons per slab ~= 4512.5 (the original result was 361,111,111 / 80k ~= 4513.88888763503 from perl -e "print 1e12/(2769.23077)/(.8*1e5)"; echo. Then, since it's a 3-d volume acquisition, take the cube root of 4513.888 to get 16.525 for the edge of the cube for monochromatic 8-bit depth data.
.
So on average, each neuronal body occupies a cube in the 3-d slab of approximately 16.5 pixels $\times$ 16.5 pixels $\times$ 16.5 pixels, if the volume acquisition is done with a single monochromatic laser using a single byte to encode intensity. If the images for the 3-d slab are acquired as an RGB dataset with three different colored lasers (which is not what is going on, i believe) then take 4513 per neuron per slab acquired and divided by three is 1504.27ish, take the cube root of that and get 11.45 pixels for the cube edge length (again assuming pixel intensity is one byte per RGB color channel). Smaller again if colors are encoded using more than one byte per color channel. Arithmetic done. (or
Maybe two decades or so ago, I recall seeing movies of scanning electron micrographs (SEM) of microchips (CMOS?) strobed with periodic inputs of fairly high frequency. By making slight changes in the input frequency, it was possible to see individual signals (electron charges) travel across the traces on the chip. I've been waiting to see this sort of visualization technique available for biological neural networks. The temporal resolution isn't quite there yet, but it appears to be coming.
I read the article and the work seems pretty good. It's important to bring up that this is a new facility and scientists there (mostly post-docs) are kept in an isolated environment and expected to work 80+ hour weeks. They've had several suicides in the last few years by these post-docs. Their director, Gerry Rubin, has acknowledged that the environment they have created will not allow many scientists to "thrive" (his word).
Assuming that we could get the positions and synaptic weights from this data, what's to stop us from making a neural network to reproduce the function of this fish brain?
Will it prove the goldfishes' memory span ?
One small step toward the technological singularity, but I don't think it will be parabolic advancement forever. It will end in a dissapointing wimper when it turns into an s-curve at some distant point in the future long before the heat death of the universe. But then again a hundred years or so ago someone said, "Everything that can be invented already has been invented". SO SINGULARITY, HEAR WE COME let us wait and see! http://rawcell.com