Japanese Scientists Claim To Reconstruct Images From Brain Data

In a world first, a research group in Kyoto Japan has succeeded in processing and displaying optically received images directly from the human brain. Here's the Japanese press release for good measure. One step closer to broadcasting your dreams? The research is due to be published today in the US scientific journal Neuron

Comments

This is NOT new

[oroborous]

A Berkeley group has already reported this in Nature using similar methods: http://www.nature.com/nature/journal/v452/n7185/abs/nature06713.html)

For the last time...

[philspear]

People read blurby summaries, which don't include the results, the full reasoning, methods, etc, and then act as if it's the fault of the researchers. It's absurd, that's neither the paper nor the direct work of the researchers, it's some non-scientist working for a news source. Read the actual paper, TFA in these cases are rarely any better than TFS.

http://download.cell.com/neuron/pdf/PIIS0896627308009586.pdf

There's the PDF. It does have the very pixelated images. I haven't had time to read through it.

As always, don't complain to me if you don't happen to have a subscription, and not having a subscription is no reason to act as if the results aren't real.

Re:No pictures?

[Peeet]

It's right there on the Japanese press release page, you can see at the bottom of the image at the top left of the article, they have the before and after of the word "neuron". Here, I'll make it even easier for ya: http://www2.asahi.com/kansai/news/image/OSK200812100099.jpg

Nope, dreams would just be noise

[Dr. Manhattan]

From what I can gather, they're pulling in rather low-level data - essentially 'listening in' on the very lowest level of pattern-recognition that's applied to the data coming in from the optic nerve. That's certainly interesting, but a whole lot more processing happens at higher levels before you 'see' anything. (C.f. people who lose sight early on due to eye problems, and have sight restored later - their brains can't do much with the information at first.)

Dreams appear to be based on the 'noise' coming in, but a lot of interpretation is applied (and without imposed constraints of consistency or logic). A common game/prank involves people asking yes/no questions about an alleged dream, but the answers they get are based on some simple scheme like "yes if the last word in the question they ask ends in a consonant". Surprisingly detailed 'stories' get constructed... by the person asking the questions. (Here's what appears to be an online version.) Actual dreams seem to be built in an analogous way, with the subconscious 'asking questions' of the senses (which are just feeding in 'static') and weaving an experience out of them.

I'd guess that 'eavesdropping' on dreams via this means would only get the kind of swirling colors and such you 'see' when you close your eyes.

Re:Pics or it didn't happen

[philspear]

Maybe this image will not require a subscription, although I suspect it will.

http://www.sciencedirect.com/cache/MiamiImageURL/B6WSS-4V4113M-P-7/0?wchp=dGLbVtz-zSkzk

On the off chance it does, keep in mind this is not the full article. Critiques along the lines of "this doesn't prove anything," or "They should have done X" are premature if you haven't read the full (journal) article. If you thought of it, they probably covered that in the article you're not willing to pay for.

Yes, it is

[philspear]

Didn't read the full article, but from the abstract

We show that these receptive-field models make it possible to identify, from a large set of completely novel natural images, which specific image was seen by an observer...

Our results suggest that it may soon be possible to reconstruct a picture of a person's visual experience from measurements of brain activity alone.

The article you linked to seems to only be able to tell which object a person saw from their fMRI. I believe it required established measurements too, IE "this part of the brain lights up when they see a face. In blind studies, that part of the brain lit up, so they must have seen a face."

Whether it required a calibration for each individual or not, no image reconstruction was done: it's not the same thing at all.

Yet Another Unnew Result

[DynaSoar]

The primary visual cortex (V1) has already been shown to be retinotopic. What's being seen can be mapped directly from the cortex. It's crude and low-res, but it works.

20 years ago a researcher working with Karl Pribram at Radford University was able to detect signals from small cellular assemblies of the visual cortex that represented a particular shape being viewed without mapping the entire shape from V1.

In both these, the images were received directly from the brain. In both they were digitally processed and presented. In all three what was retrieved was not an image, but was a pattern of neural electrical activity that they had already determined represented a particular visual field. They could not (in keeping with the /. tendency to represent reality with fiction) for instance, retrieve the third frame of a series of images that had been briefly presesnted. They would have had to show the image for some time that record EEG from the appropriate areas for long enough that they could get a good correlation when showing it a second time.