News:

NOTICE: Posting schedule is irregular. I hope to get back to a regular schedule as the day-job allows.


Monday, August 27, 2012

NEWS: Hacking the Brain

http://teddysratlab.blogspot.com [Full link to blog for email clients.]


The recent report from UC Berkeley about a "new" techniques for "hacking" the brain (http://seattle.cbslocal.com/2012/08/25/scientists-successfully-hack-brain-to-obtain-private-data/), is neither new, nor is it hacking.  The research relies on recording what neuroscientists call the "P300" brain wave.  To explain why this is important (but not new), let me digress a moment..

Copyright 2012 by "jimmi", Licensed by Shutterstock
The EEG is a continuous readout of electrical activity that is affected by all activity of the brain at a given instant.  Because the brain is constantly active and processing information, it is difficult to point to any single peak and understand the information it represents.  Neuroscientists get around this issue by repeating events – sights, sounds, touches – then averaging the EEG synchronized to those fixed points in time.  EEG peaks that always occur in response to an event will continue to show up in the average, while "random," or at least asynchronous, events get averaged out.  The term for these patterns in averaged EEG is "Event-Related Potentials" and they have been studied since the 1930's, and the "P300" has been studied since the 60's.

AS an aside, ERPs are how doctors track visual and hearing deficits in infants.  The ERP peaks that occur within 100 milliseconds of a stimulus corresponding to information processing by various stages in the visual and auditory pathway.  "Counting the bumps" lets a physician determine whether (and where) a deficit may occur.  After 100 ms, the ERP peaks corresponding to different features – most notably the concepts of "expectancy" and "surprise".  IF you flash or play stimuli to a subject at a fixed rate, and omit one stimulus in a chain, such that the subject *expects* the stimulus, but it does not occur, then a large negative peak at 100 ms (N100) occurs.  If an unexpected or rare event occurs, a large positive peak occurs in or around 300 ms – hence the "P300" designation. 

P300 is more properly termed "P3" since it actually occurs from 250-400 ms (depending on the stimuli and species), but is the third major positive peak in the ERP.  It is also split into two components – P3a (250-300 ms) which responds to truly novel stimuli, and P3b (300-400 ms) which responds to rare, infrequent yet familiar stimuli.  It is very important to note that P3b is only elicited within the context of the subject's attention.  The most famous test of P3 is to tell a patient to count high pitched tones, then throw in one low pitched tone every 10 high-pitched tones.  The "oddball" stimulus elicits a large P3b, and the lower the frequency of "oddball" stimuli, the higher the peak of the P3b.  If, on the other hand, you throw in a flash of light instead of a tone, a P3a occurs and not a P3b since the stimulus was novel and not relevant to the task you assigned the subject.

P3 is not knew, by any means.  The earliest reports are from the 60's, and most of the "oddball" tests date from the late 70's and early 80's.  In fact, my PhD dissertation from the mid 80's was on tracing a P3-like phenomenon in rat brain that predicted sequences of stimuli and actually biased the animal's future behavior (the first step in my current research which predicts and even controls behavior via a neural prosthetic). 

So P3 is not "new", nor is it hacking.  Since the 80's P3a and P3b – and in fact, the whole set of ERPs – has been examined for use in lie detection, but there are various interpretations – the first is that a "lie" would result in a P3b due to "guilty knowledge."  Subjects are examined by reading back their answers to given a series of questions, and supposedly hearing their "lie" would result in a large P3b since it conflicts with what they know to be true.  The fallacy is that for ERP recording to be reliable, the EEG must be averaged, thus the Q&A needs to be repeated enough times to average out signals not specific to the stimuli, the "lie" is no longer rare or unexpected, thus a P3b is less likely to be generated.  In addition, a well-practiced lie becomes the more familiar answer, and is thus more likely to produce a smaller P3b, which in turn is less likely to be distinguished from truthful, familiar answers. 

The second reason why use of P3b is not "hacking" is that the procedure is unlikely to work quite the way it is described in the article.  The press release states that the procedure could be used to discover PIN numbers and passwords because the subject generates a "P300"  when they see or hear their secret codes presented.  IN fact, since the codes are private, yet highly familiar, it is much more likely that the combination of N100, P300 and other ERPs that respond to novelty and unfamiliarity will be smaller when their own PIN or password are presented.  The procedure may work as described in some cases, but it is highly unlikely to work as described in subjects who are aware of what the procedure is attempting and have those codes extremely well memorized.

I love seeing articles like this because they provide an opportunity to raise awareness of brain science – but at the same time I hate them because university press offices try for a spin that is the most likely to catch public attention irrespective of the dull truth of the science.  In the past I have blamed the press offices and not the scientists when they misstate the science.  Given that I am currently working on press releases on my own work, I can see how easy it is to go with the simple, spectacular pronouncements, but I do try to keep them within the range of scientific accuracy and clearly indicate the speculative bits.  There is still a measure of truth in the report, especially in that ERPs can and do provide an important measure of how the brain is reacting to and processing information.  ERPs are also very important in detection and analysis of disease states and can provide valuable tools for brain-to-computer interface for locked-in patients.

As always, I caution my readers about over-interpretation of publically-released scientific information, and certainly invite inquiries, as well as refutations, as they have provoked many interesting debates and provide an opportunity for learning.

No comments:

Post a Comment

Please add comment - no links, spammers will be banned.