fMRI 101

63 views
Skip to first unread message

Ivan

unread,
Sep 24, 2011, 11:43:13 PM9/24/11
to HackerspaceSG
So, another fMRI-based mind-reading study [1] made it into slashdot,
hacker news and related news feeds. And there's a worrying "2.0"
component in it...
How many people would be interested in learning about fMRI basics,
what's cool and uncool about it, before rushing to pay 30 bucks to
Elsevier to read the actual article?

[1] http://newscenter.berkeley.edu/2011/09/22/brain-movies/

Tamas Herman

unread,
Sep 25, 2011, 1:42:49 AM9/25/11
to hacker...@googlegroups.com
i would be interested

--
tom

Patrick Haller

unread,
Sep 25, 2011, 5:02:49 AM9/25/11
to hacker...@googlegroups.com
On 2011-09-25 11:43, Ivan wrote:
> So, another fMRI-based mind-reading study [1] made it into slashdot,
> hacker news and related news feeds. And there's a worrying "2.0"
> component in it...
> How many people would be interested in learning about fMRI basics,

As long as someone makes a Scanners [1] reference ;)


Patrick


[1] http://www.imdb.com/title/tt0081455/

Jasper C

unread,
Sep 25, 2011, 8:46:09 AM9/25/11
to HackerspaceSG
er, I could probably give a basic intro to fMRI. And is there a link
to the Elsevier article? Or a title?

Actually reconstructing an image out of fMRI scans sounds so far
advanced I'm tempted to call horse pucky...

Jasper C

unread,
Sep 25, 2011, 10:45:16 AM9/25/11
to HackerspaceSG
Doh! Should have said reconstructing an image of what the subject
sees... fMRI images are pretty standard. :P

Ivan

unread,
Sep 25, 2011, 11:38:50 AM9/25/11
to hacker...@googlegroups.com
Hi Jasper,

On Sun, Sep 25, 2011 at 10:45 PM, Jasper C <jaspe...@gmail.com> wrote:
> Doh!  Should have said reconstructing an image of what the subject
> sees...  fMRI images are pretty standard. :P

yeah, i don't think we need to get into k-space here...

>> er, I could probably give a basic intro to fMRI.  And is there a link
>> to the Elsevier article?  Or a title?

http://www.sciencedirect.com/science/article/pii/S0960982211009377

Couple of years ago there was a report from a Japanese group doing
something similar (guessing (visually presented) words from visual
activation areas) (don't have the reference right now, but can dig it
out). It was also followed by
imminent-mind-reading-technology-availability-OMFG in popular press.


Actually, I don't want to talk about these particular studies. I find
it interesting that fMRI (BOLD) is still considered new/cool/emergent
technology after 15+ years. So, exposing the basics and limitations of
the technology will potentially help some people to react
appropriately to overzealous journalists and/or neuro-marketers...

cheers,
Ivan

Benjamin Scherrey

unread,
Sep 26, 2011, 4:25:14 AM9/26/11
to hacker...@googlegroups.com
I was thinking Brainstorm ( http://www.imdb.com/title/tt0085271/ ) with Christopher Walken and Natalie Wood (who died just after this was filmed). Just inspired to find a copy as I haven't seen it in about 15 years. I remember it being very well written with great actors and amazing videography. I bet this makes my Top 20 at least.

  -- Ben

Jasper C

unread,
Sep 26, 2011, 8:26:32 AM9/26/11
to HackerspaceSG
Right, so had a look at the paper. Just a brief one because it's one
of those papers that tries to use as many big words as possible...

So yes, they are reconstructing images of what the subject sees based
on fMRI scans. There are a few caveats: They first need to collect
data - 2 *hours* of fMRI scans with the subject watching a training
set of data. And then they test with 9 minutes of test data. There's
also mention elsewhere that they need a unique model to each subject -
not mentioned if that's generated from the 2 hours of scan data or
something additional.

I asked a neuroscientist - apparently the human visual cortex is a
honking big chunk of brain, so it is possible to gather enough data to
do it.

Apparently the technique itself isn't new. The new bit is that
they've found a way to get round the fact that hemodynamic process
happen slowly. Slowly enough to make this ordinarily unfeasible.



On Sep 26, 4:25 pm, Benjamin Scherrey <proteus...@gmail.com> wrote:
> I was thinking Brainstorm (http://www.imdb.com/title/tt0085271/) with
> Christopher Walken and Natalie Wood (who died just after this was filmed).
> Just inspired to find a copy as I haven't seen it in about 15 years. I
> remember it being very well written with great actors and amazing
> videography. I bet this makes my Top 20 at least.
>
>   -- Ben
>

Melissa Lewis

unread,
Sep 26, 2011, 1:00:57 PM9/26/11
to HackerspaceSG
First, they tested subjects with 9 minutes of test data repeated 10
times. Then they ultimately divided that into individual minutes from
the scenes to try to account for within-subject changes like ...
boredom.

Also, the 2 hours of data they gathered to train the algorithm for
each person were unique. That's a lot of training data! And the amount
of data gathered for a subject is a function of the area being
measured *and* time *and* how that time is divided. So while the
section of occipital cortex they measured being large is a sufficient
condition for a large data set, it isn't necessary.

Also, yeah, fMRI isn't a new technology and I don't believe it's the
data-gathering side that people are excited about here; it's the
algorithm used to interpret and represent that data. All that said,
I'm also looking forward to the day when fMRI research is treated in
popular media with the caution it deserves, but I don't think the
popular media are entirely to blame. Those scientists whose focus is
using sexy techniques do so because sexy techniques get published, and
a disproportionate level of attention is afforded them (the
techniques, that is, though I s'pose the same could be said of the
scientists). On the topic of debunking, have y'all seen the dead
salmon study? It's awesome: http://neuroskeptic.blogspot.com/2009/09/fmri-gets-slap-in-face-with-dead-fish.html.

Let me know if you'd like some literature and/or videos on the fMRI
technique and its drawbacks! And for that matter, let me know anytime
you'd like to get your hands on some primary literature; I have free
access to it. And even if I did not, you can usually write to one of
the authors of a paper and simply request it!

Benjamin Scherrey

unread,
Sep 26, 2011, 1:05:04 PM9/26/11
to hacker...@googlegroups.com
c'mon Melissa - that was a ZOMBIE fish! :-) Great write up.


Ivan

unread,
Sep 27, 2011, 4:17:49 AM9/27/11
to hacker...@googlegroups.com
> Apparently the technique itself isn't new.  The new bit is that
> they've found a way to get round the fact that hemodynamic process
> happen slowly.  Slowly enough to make this ordinarily unfeasible.

I doubt they've found a way to get around this. Then again, the speed
of the response is not really an issue if the goal is to associate a
signal pattern with another one. A more significant issue with these
kind of studies is the fact that hemodynamic response is the same for
neuronal excitation and inhibition.

- Ivan

Ivan

unread,
Sep 27, 2011, 4:38:26 AM9/27/11
to hacker...@googlegroups.com
Thanks Melissa. I'd like to have a look at the original article.

To get anything useful out of data from subjects spending 2+ hours in
an MRI scanner requires some serious kung-fu.

> Also, yeah, fMRI isn't a new technology....

Yes, I made a mistake, it's already 20+ years, older than "www"...

> On the topic of debunking, have y'all seen the dead
> salmon study? It's awesome: http://neuroskeptic.blogspot.com/2009/09/fmri-gets-slap-in-face-with-dead-fish.html.

Yes, quite funny. Behind the scenes is also a good read
http://prefrontal.org/blog/2009/09/the-story-behind-the-atlantic-salmon/

Too bad that salmon dataset is probably less than 9 min long.
Otherwise I would be ready to contribute money to learn about youtube
preferences of a dead fish.

- Ivan

Jasper C

unread,
Sep 27, 2011, 11:28:55 AM9/27/11
to HackerspaceSG
> Let me know if you'd like some literature and/or videos on the fMRI
> technique and its drawbacks! And for that matter, let me know anytime
> you'd like to get your hands on some primary literature; I have free
> access to it. And even if I did not, you can usually write to one of
> the authors of a paper and simply request it!

Thanks for the offer. For the moment I have free access too. :)

And the salmon one was great. Pity it didn't get published. er, did
it eventually?




On Sep 27, 4:38 pm, Ivan <van...@gmail.com> wrote:
> Thanks Melissa. I'd like to have a look at the original article.
>
> To get anything useful out of data from subjects spending 2+ hours in
> an MRI scanner requires some serious kung-fu.
>
> > Also, yeah, fMRI isn't a new technology....
>
> Yes, I made a mistake, it's already 20+ years, older than "www"...
>
> > On the topic of debunking, have y'all seen the dead
> > salmon study? It's awesome:http://neuroskeptic.blogspot.com/2009/09/fmri-gets-slap-in-face-with-....
>
> Yes, quite funny. Behind the scenes is also a good readhttp://prefrontal.org/blog/2009/09/the-story-behind-the-atlantic-salmon/

Gideon Simons

unread,
Sep 28, 2011, 10:53:04 PM9/28/11
to hacker...@googlegroups.com
I wonder how they get the bugdet for all this, MRI experiments are so expensive.
I think the greatness of this article is the whole youtube database brain activity to imagery matching, looks like a tool with potential.
But I think the whole sexyness and revolutionary headline is a bit miss leading to the fact that its not really a visual pattern that is reconstructed from the brain activity its just a blurred out youtube clip scene that what the best match. The video at first glance makes it even more miss leading.

The whole idea and experiment is really interesting for what it really is, I think it may be a great tool to use to validate other experiments like emotional reactions to visuals etc.. and maybe one day people will be able to upload their dreams reconstructed out of youtube scenes to youtube :) .

If you want to see a really interesting experiment using MRI, here is probably the least sexy but most revolutionary one that I have seen:

Gideon








--
Gideon

Denisa Kera

unread,
Oct 4, 2011, 10:32:22 PM10/4/11
to HackerspaceSG
In May I talked to people from Prague based Brmlab about this
http://brmlab.cz/project/brain_hacking
They can help if we need any info, support etc., they try to share
their data with the science community but without much effect,
interesting research into lucid dreaming & sleeping, they can induce
lucid dreaming via electrodes placed at the right place on the
head...
In Paris there is another Hackerspace (don't have links only e-mails)
that does a lot on brain-hacking & sonification, they use the brain
data to make music over pure data and similar software, I really liked
that.
I'm not that interested in brain issues but can connect you with these
people that actively experiment & I'm definitely curious to see what
happens.
Reply all
Reply to author
Forward
0 new messages