Is a bug in FMRI software causing decades of brain research to be invalidated?

[gva]

http://www.sciencealert.com/a-bug-i...decades-of-brain-research-scientists-discover

Could this be true?

 

[Grinkle]

My .02 after reading -

That image processing software has bugs should not be surprising to anyone who has visibility into the realities of complex software development and maintenance.

The speculation that 15 years of research is potentially invalid is just that - speculation. My own speculation is that the reality is much more grey than black and white.

The false positive rate seems pretty high to me, so to the extent that the specific detection mechanism cited is central to a previous conclusion, I would think that conclusion is at risk of being not-duplicatable or perhaps weaker than initially thought. My guess is that the principle of brain mapping is sound, the existing maps are probably noisier than they will become as mapping techniques continue to improve.

I am certain that the last bug in imaging software has not been found, its a journey.

 

[Ygggdrasil]

It is certainly believable. fMRI studies have been notoriously unreliable as a research technique and subject to many false positive. For example, to demonstrate the ease of discovering false positives, researches placed a dead salmon from the supermarket into an fMRI scanner and, using the same analysis techniques used in many published papers, found signs of activity in certain regions of the brain when showing the salmon pictures of human faces. Furthermore, fMRI is a very indirect measure of brain activity, measuring things like increased blood flow or increased metabolic activity in different areas of the brain. To what extent these observables correlate with the underlying activity of neurons is still unclear.

Here's a nice piece from a few years ago highlighting some of the issues with fMRI and other similar brain scans:
http://blogs.scientificamerican.com/guest-blog/controversial-science-of-brain-imaging/

For the above reasons, I've never put much faith in the 15 years of brain research that has been invalidated in the first place (but I'm not a neuroscientist). Understanding how the brain works is difficult, in part, because we lack good tools for measuring and monitoring brain activity. This is certainly an area where new technologies would be very helpful (the impetus for Obama's BRAIN initiative).

 

[DiracPool]

I have high doubts that this is a significant issue, despite Ygggdrasil's alarmism. Don't believe that 40,000 plus research reports are going to be flushed down the drain. There's a remarkable amount of consistency in the data results in extant "resting-state" fMRI brain literature that comports with knowledge of the evolutionary and functional neuroanatomy of the brain. The likelihood that a computer glitch is going to confound BOLD signal readings in a significant portion of the 40K papers in question is vanishingly small. So I wouldn't lose sleep over it. That doesn't mean that the software doesn't need to be up to standard, though, far from it, it just means that don't be too quick to throw out the baby with the bathwater.

Ironically, part of my job as a research assistant in an AI lab this fall is going to be to do data management on the 1000's of fMRI studies out there. I'm sure the head of the lab is already aware of the media reports presented here, but I'll bring it to his attention, anyway. I'm actually going to meet up with him in a about a week and half at the IEEE WCCI in Vancouver. So I'll give you the updated report on the matter..

http://www.wcci2016.org/index.php