From phys.org we learn that common statistical methods which are used to analyze brain activity images taken with fMRI (Functional Magnetic Resonance Imaging) scanners may not be trustworthy. From the article:
[Anders] Eklund [...] tested the analysis methods by using them on known, reliable data. The methods showed false activity in the brain in 60 percent of the cases. A reasonable figure is five percent.
The analysis methods were previously only validated against simulated data, but Dr Eklund pointed out back in 2012, in his doctoral thesis, that the results are not always reliable. The statistical methods used are built on a number of assumptions; if one or more of the assumptions are incorrect, the results will also be incorrect. At that time, critics maintained that errors could certainly arise if data from a single person was analysed, but that the errors would even out in group analyses.
In his thesis, Dr Eklund proposed another method in which few assumptions are made and significantly more calculations - a thousand times more - are done, which yields a significantly more certain result. With the help of modern graphics cards, the processing time can be reduced so that the method is usable in practice.
[...] Dr Eklund now used the current analysis methods and compared 20 healthy people with 20 other healthy people. In other words, there should not have been any differences - or, in any case only the five percent that chance provides. In total he made three million comparisons of randomly selected groups with data from 499 healthy persons.
"The differences were considerably greater than five percent, up to 60 percent in the worst case," Dr Eklund says.
This means that the analyses could have shown positive results where there shouldn't have been any, thereby indicating brain activity where there was no activity.
He also analysed the same data set with his more calculation-heavy method and obtained a considerably better correspondence, with differences in the expected five percent of cases.
The paper will appear in Proceedings of the National Academy of Sciences. A preprint is available on ArXiv.
(Score: 1, Insightful) by Anonymous Coward on Tuesday June 28 2016, @08:35PM
(Score: 2) by fork(2) on Tuesday June 28 2016, @08:59PM
You're absolutely right. Apparently both instances of "fMRI" were altered while I was editing the article :(
(Score: 2) by TGV on Wednesday June 29 2016, @07:37AM
fMRI *is* MRI. It's the way it's applied and the data is interpreted that makes it "functional".
(Score: -1, Disagree) by Anonymous Coward on Tuesday June 28 2016, @08:41PM
People really do believe this nonsense:
There should not be any difference between two groups of healthy people?
(Score: 0) by Anonymous Coward on Tuesday June 28 2016, @09:11PM
No, there shouldn't be, since in each case you're averaging the entire group of twenty. Individual differences average out, leaving the common-mode signal; this should indeed be the same. (Of course there's an unstated requirement that you divide the forty samples into two groups of twenty at random, rather than sorted by some parameter, but duh.)
(Score: 0) by Anonymous Coward on Wednesday June 29 2016, @05:40AM
Yes, you are correct it actually refers to two groups sampled from populations with the exact same distribution, or sometimes distributions with the same mean (it all depends on the test). Doesn't matter.
You will be the first to ever explain it. How exactly does randomly allocating people to different groups ensure they are the same on average regarding whatever you are measuring?
(Score: 0) by Anonymous Coward on Wednesday June 29 2016, @03:55AM
fMRI are used to measure blood flow, which is then used to infer brain activity, it does not measure brain activity directly. That it is error-prone to sometimes show areas might be active when they're not is no surprise.
(Score: 2) by fleg on Wednesday June 29 2016, @07:55AM
>fMRI are used to measure blood flow, which is then used to infer brain activity, it does not measure brain activity directly.
no.
from wikipedia [wikipedia.org] ...
>That it is error-prone to sometimes show areas might be active when they're not is no surprise.
no.
tfa is about an error in the underlying assumptions of "common statistical methods which are used to analyze brain activity images taken with fMRI".
i aint no maffs head by any stretch but that error seems to be that they assumed a gaussian distribution where there wasnt one.
(Score: 2) by JoeMerchant on Wednesday June 29 2016, @12:03PM
fMRI is weak magic, inference on top of heavy statistics on top of several assumptions. When they were first published it was more of a science fiction "wow, look at this single subject, wouldn't it be awesome if this actually worked" kind of thing, but over the years many researchers in other fields have warmed up to the KoolAid, or completely ignored the caveats, and started drawing conclusions - always couched in a "well, anatomy varies from one subject to the next, so that explains the wild data distributions we're seeing."
Even this study doesn't really satisfy... I'd be much more impressed with 100 animals both fMRIed and independently measured by some other method, but that would be "too expensive" and possibly cruel, so let's just continue publishing papers about studies done on groups of 5 humans and taking up everybody's time and attention with mostly baseless wild conjecture.
🌻🌻 [google.com]