MRI software bug invalidates 40,000 research papers

Please consider donating to Behind the Black, by giving either a one-time contribution or a regular subscription, as outlined in the tip jar to the right. Your support will allow me to continue covering science and culture as I have for the past twenty years, independent and free from any outside influence.

The uncertainty of science: A bug just discovered in the computer software used by MRIs to measure brain activity could invalidate 15 years of research and 40,000 science papers.

They tested the three most popular fMRI software packages for fMRI analysis – SPM, FSL, and AFNI – and while they shouldn’t have found much difference across the groups, the software resulted in false-positive rates of up to 70 percent. And that’s a problem, because as Kate Lunau at Motherboard points out, not only did the team expect to see an average false positive rate of just 5 percent, it also suggests that some results were so inaccurate, they could be indicating brain activity where there was none.

“These results question the validity of some 40,000 fMRI studies and may have a large impact on the interpretation of neuroimaging results,” the team writes in PNAS. The bad news here is that one of the bugs the team identified has been in the system for the past 15 years, which explains why so many papers could now be affected. [emphasis mine]

The research the article described is focused entirely on the problems the software causes for past research. It makes no mention of the problems this software bug might cause for actual medical diagnosis Was the treatment of any patients effected by this bug? It does not say.


  • Max

    Wow!… Faults positives up to 70%? I wonder if the brochure from the MRI salesman said that they guarantee up to 70% more patients?
    Now we know why Canada has so few MRI machines. (I wonder if their software maker is the same company that runs our voting computers…)

  • wayne

    in all seriousness however, the reason we have a lot of MRI machines in the USA, is/was primarily the free-market.
    20 years ago it was perfectly legal for Doctors to buy/invest-in, their own MRI machines, load them on tractor-trailers, and run them almost 24/7. (often, in the parking lot of their office’s)
    In the town where I lived at the time, two medical groups did exactly that & literally years before the local Hospital received “permission” to put one in.
    Various “health-care reforms,” primarily in the last 8 years, made that illegal. (It’s a “conflict of interest,” apparently & we need to be “saved” from “exploiting doctors.”)
    -Fortunately, we have a huge installed base of MRI machines (& CT scanners) already in place. I seriously doubt we would be in that position, if MRI machines were invented today, and controlled under the ACA.

    As for the software errors; I know just enough Statistics to be dangerous, but that being said– if the error is consistent across time, software, & machines, are there not mathematical ways to extract that out, to a least some degree? and maybe not invalidate every single study??

    Tangentially– personally, I’m highly suspect of the “active” real-time MRI scans they tout as being evidence for/against someone’s theory of brain-activity, but that’s a different topic above & beyond inherent software errors.

  • Interesting post. I agree with you on the last thing you mentioned in your comment above, regarding the “active real time MRI Scans”

Leave a Reply

Your email address will not be published. Required fields are marked *