Groundbreaking 3D brain scan generated 1.4 petabytes of data from millimeter-sized sample

zohaibahd

Posts: 91   +1
Staff
What just happened? Researchers have reconstructed a minuscule piece of the human brain down to the level of individual synapses, representing a giant leap forward for brain science. And we're not just talking about a few neurons here. This millimeter-sized cube contains a staggering 57,000 cells, 230 millimeters of tiny blood vessels, and nearly 150 million synaptic connections, all mapped out in glorious 3D detail.

The process began with a surgically removed sample of brain tissue taken from a woman with epilepsy during brain surgery meant to help control her seizures. After chemical preparation to enhance contrast, the specimen was embedded in resin and sliced into an astounding 5,000 sections, each about a thousandth the thickness of a human hair.

From there, high-throughput electron microscopy was used to scan each sliver, generating a staggering 1.4 petabytes of raw data. The Google team then used a machine-learning model to align and reconstruct all the 2D images into a detailed 3D dataset.

The map contains the highest-resolution picture of the human brain ever created, covering a cubic millimeter that is a millionth the size of the entire brain.

The math whiz kids over at Tom's Hardware performed some number crunching to calculate the space the entire map of the human brain would need. That would add up to 1.6 zettabytes of storage – a data center costing $50 billion and spanning 140 acres.

The model has already revealed some surprising insights into brain architecture at the cellular level. For one, the non-neuronal cells that support and insulate neurons outnumbered them by around 2-to-1, with oligodendrocytes being the most prevalent type, forming the protective myelin coating around axons.

On average, each neuron made connections with thousands of other neurons. But the team also identified rare instances of single axons forming over 50 high-powered synaptic connections to the same neuron, as well as some axons coiled into long, intertwined "whorls" for reasons not yet clear.

Since the sample came from an epilepsy patient, some of these quirky structures could be related to the disorder rather than normal brain anatomy.

Regardless, the project has the potential to expand our understanding of the brain's incredible architectural complexity at a cellular scale. After all, there's still a staggering amount we don't know about how the human brain's hardware operates.

The map can be freely accessed by researchers from a web platform called Neuroglancer for their own studies into how the human cortex processes information and stores memories at this unprecedented resolution.

Permalink to story:

 
A staggering amount of data, goes to show we could be several decades before brain implants and things are practical. Imagine the size of the computer that would be needed to store a human brain in memory process what the brain is doing in real time.

I've been wondering what massive computers could possibly be used for outside of research and basically AI. It's interesting to think of the practical applications of that kind of computer power. It also shows we aren't nearly as advanced as we like to think we are.
 
The obvious question would be to what end would this be used? AND how can governments use it against their own citizens? Privacy is and will continue to be a thing of the past .....
 
So how long until we can develop a personal device that can store the entirety of our brain - and maybe our close family members’ as well?

Imagine a future where instead of your medical file being just your past appointments and medical tests, but all of your personal body data (brain, liver, heart etc).
 
So how long until we can develop a personal device that can store the entirety of our brain - and maybe our close family members’ as well?

Imagine a future where instead of your medical file being just your past appointments and medical tests, but all of your personal body data (brain, liver, heart etc).
If I ever can upload my brain into a computer I can promise you no one else is coming with me
 
It is interesting, as a lab test, but practically it doesn't tell us much. The number of neurons and synaptic connections in human brain has been evaluated years ago, so those numbers are of no surprise. And as far as brain scanning goes, cutting out pieces from it is not a viable solution for modeling the brain as a whole. It would be more interesting to scan the entire brain, even while extracting only partial information, as this would let us collect a lot of interesting statistics about types and complexity of all connections.

Since the sample came from an epilepsy patient, some of these quirky structures could be related to the disorder rather than normal brain anatomy.
I'm not a brain specialist, but I'm still willing to stick my neck out, and call it a nonsensical assumption, as much as suggesting that epilepsy produces a new brain structure.

In all, a real progress would be to scan the entire brain, but initially without storing any information, only statistics - ranges of complexity and types of connections. Such data would be very useful in understanding the scope of structural differences and let compare brains that way between people. Again, no need of storing any large data, just a statistical summary.
 
Last edited:
I'm not a brain specialist, but I'm still willing to stick my neck out, and call it a nonsensical assumption, as much as suggesting that epilepsy produces a new brain structure.
It's not an assumption, it's what science does: admit what we don't know. Scientists studying this scan will keep in mind that epilepsy could play a role in the structure they are observing. Will they actually observe anything caused by epilepsy? We won't know until other brains are compared.

It is interesting, as a lab test, but practically it doesn't tell us much.
Not yet. Science at this scale and level of detail takes time.

In all, a real progress would be to scan the entire brain, but initially without storing any information, only statistics - ranges of complexity and types of connections.
It would be interesting, though the scanners and compute calculating those statistics would still have to churn through the 1.6 zettabytes of raw information (just wouldn't be storing it), probably not a trivial feat (and one would need a donor brain to do it). The statistical summaries would have to be detailed enough, and they would only be at a per-slice level (since we aren't keeping the scans themselves around to stitch them together), so connecting the statistics from each slice together would be a separate exercise. Still, until storage at the zettabyte scale is cheaper, this isn't a bad idea if the scanning time is reasonable (and it may not be).
 
So how long until we can develop a personal device that can store the entirety of our brain - and maybe our close family members’ as well?

Imagine a future where instead of your medical file being just your past appointments and medical tests, but all of your personal body data (brain, liver, heart etc).
An interesting idea, would probably have to do all of these things:
Has as high or higher storage density than the brain (if you want it to be something that can be carried)​
Has a reasonable power requirement (the brain consumes about 20 watts, though the device could consume more)​
Can be synchronized with the brain (would need a high bandwidth, perhaps on the petabit/second level if not greater, depending on what is being synchronized and how often, and high compute, to make the necessary updates)​
Would need to operate without interfering with the brain​
Might need to interact with the peripheral nervous system (if gathering more body data)​

In short I don't think this technology will be around for a very long time, maybe not even in the 22nd century. If you ignore the miniaturization requirements around the device being carried, it could be possible then, but it would still be quite the feat. Not to mention getting folks to agree to it, though with the rise of AI technologies, this kind of technology might be the kind of thing that keeps humans relevant.
 
An interesting idea, would probably have to do all of these things:
Has as high or higher storage density than the brain (if you want it to be something that can be carried)​
Has a reasonable power requirement (the brain consumes about 20 watts, though the device could consume more)​
Can be synchronized with the brain (would need a high bandwidth, perhaps on the petabit/second level if not greater, depending on what is being synchronized and how often, and high compute, to make the necessary updates)​
Would need to operate without interfering with the brain​
Might need to interact with the peripheral nervous system (if gathering more body data)​

In short I don't think this technology will be around for a very long time, maybe not even in the 22nd century. If you ignore the miniaturization requirements around the device being carried, it could be possible then, but it would still be quite the feat. Not to mention getting folks to agree to it, though with the rise of AI technologies, this kind of technology might be the kind of thing that keeps humans relevant.
It's always dangerous to predict future tech solely on the past - even if that's the easiest way to do it... the human mind is quite often capable of exceeding these predictions... ask some of the leading scientists who developed the first computers just a few decades ago if we'd be capable of what we have now and they'd probably have scoffed...

Nature has found a way of storing far more data inside something weighing about 3 pounds - that would be our brain... and the power requirements aren't really that high... I don't think it's that far-fetched to believe that our leading scientists will surpass "mother nature" at some point - they seem to do it all the time in other fields.
 
It would be interesting, though the scanners and compute calculating those statistics would still have to churn through the 1.6 zettabytes of raw information (just wouldn't be storing it), probably not a trivial feat (and one would need a donor brain to do it). The statistical summaries would have to be detailed enough, and they would only be at a per-slice level (since we aren't keeping the scans themselves around to stitch them together), so connecting the statistics from each slice together would be a separate exercise. Still, until storage at the zettabyte scale is cheaper, this isn't a bad idea if the scanning time is reasonable (and it may not be).
Today's supercomputers can churn through multiple terabytes a second with ease. And there is no shortage of donors for this kind of thing. There is only one challenge - the scanner, to be able to scan that amount of data, within reasonable time, no other problems really. One of the biggest challengers about scanners is that increasing its power for better scan can affect the brain, compromise scan integrity, and worse - damage it (while a weaker scanner lacks in speed and penetration).
 
Once they find new ways to scan and use more advanced tech, it will be more.
I recall something in gigabyte range when scientists began to store data from human body.
Their data required gigabytes when gigabytes were the biggest unit of storage.
But that was much earlier when terabyte hard drives did not exist.
It will be much more than petabytes which is understandable
when we talk about the most advanced living being on earth.
 
Back