Slashdot Mirror
Scientists 'Read Thoughts' Using Brain Scans
Bruce_of_the_Cosmos writes "Researchers at University College London and University College Los Angeles say that the can 'read' thoughts using fMRI brain scans. While a subject's attention switched between two images, scientists could monitor activity in the visual cortex and accurately determine, among other things, which image the patient was looking at."
They couldn't guess because it wasn't a multiple choice. The summary says that they need the person to look at different choices, and they can determine which choice has been chosen. So it's not really a useful interrogation means - yet, unless they show the suspect a list of pictures and can objectively determine what thoughts each picture invokes. And I have a feeling it would be a lot less accurate if the participant isn't a willing subject.
Saskboy's blog is good. 9 out of 10 dentists agree.
The time course of fMRI is currently way too slow for use in neuroprosthetics. As for reading thoughts -- the studies looked at primary auditory and primary visual cortex, the two cortical areas least likely to be involved in conscious thought. The mind reading, neuroprosthetic spin is just that, spin. The really importart finding in these studies is the correlation of fMRI signals with electrical activity in the brain. fMRI measures increases in blood flow which has been suggested to be caused by increases in electrical activity in the brain - these studies provide evidence to suport this hypothesis. Scientist that study the electical signals in the brain directly (like me) have routinely critized fMRI studies because until now in was unclear how the results related to signal processing in the brain. There is still one major short coming of fMRI. Imagine that 50% of the neurons in an area of the brain increase their electrical activity while 50% equaly decrease their activity. This would result in a large change in signal processing but no change in blood flow and therefore would not show up in a fMRI scan. That said, fMRI is a powerful tool for understanding neural function, particularly in human who for some reason object to letting you stick electrodes into their brains. These new studies make in an even more useful tool.
My research work (and my doctoral dissertation) involved developing technology to enable exactly these studies. The basic mechanism which these studies use was published back in 1992 by three groups almost simultaneously (Harvard-MGH, U. of Minnesota and the Medical College of WI).
After almost 15 years, the workings of the brain that causes this phenomenon is still not completely understood. What happens when a region of the brain starts working towards a particular mental task, be it visual, auditory, memory, etc., is that blood supply to that part of the brain increases to such an extent that there is an oversupply of oxygen (via hemoglobin). The differing levels of oxy- and deoxy- hemoglobin have different enough magnetic properties that the change in relative amounts can be detected by a suitably equipped MRI scanner.
I've been telling this joke at parties for years when people ask me what I do - much better than saying I'm an engineer developing MRI hardware and software.
Bottom line - we've been able to do this for years. But the workings of the living brain are incredibly complex, and it'll be a little while before we get to the bottom of things. That piece on lie detectors using brain scans that came out a few months ago was based on this same technology/research. But we really don't know anywhere near enough for me to think that research was anything close to valid.
Every time someone publishes an interesting fMRI result the press call it mind reading. This is study about binocular rivalry and being able to predict which of two rivalrous stimuli are being attended just by looking at MR signal. Lots of people are working on this sort of thing. What happens is that under certain conditions, when two stimuli are presented separately to each eye, rather than combining the images, the brain maintains both separately and "switches" between which of the two are currently being attended. You have some limited ability to control which of the two you attend to, although you kind of habituate and then spontaneously switch. It's similar to viewing a Necker cube: you can switch which faces are in front or in the back of the cube. Anyway, the coolness of this study is that they could tell which of the two stimuli were being attended just by looking at the brain data (confirmed by the subjective reports of the participants in real time). It's important to note that they don't do this in real time! The MR data take a lot of post-processing and statistical analysis before they get anything out of it.
Anyway, the novelty here is that rather than stimulus predicting what brain area should be recruited (like most MR vision studies), they say, given that this bit of brain lit up, we're going to predict what you were looking at (or in this case, attending to). This is mind-reading, but you know, only in the most academic and post-hoc sense. It's not the first time it's been done, btw. Jim Haxby has done this sort of thing with people looking at overlapping pictures of people and places.
It's cool (to scientists) without needing to sensationalize it as mind-reading. Real mind-reading is coming, don't worry. But not for decades, if not a century. And yes, the government is interested in it (they approach brain scientists about this sort of stuff all the time). Right now they want a "better" lie detector. (By which, I suppose, means one that works at all since the polygraph is bunk). But we're a long, long way off.
What you're seeing is not the raw data.. we don't know how to get that (if 'raw data' even makes sense in this context).. you're seeing noise from the activity of the brain. It's like sticking an aerial next to a PC, recording the RF noise and playing that back next to a different PC - you aren't going to suddenly get a free copy of Quake downloaded into it...