News blog

Study warns of red herrings in brain scan data

Posted on behalf of Kerri Smith

<img alt=“fish fmri.jpg” src=“http://blogs.nature.com/news/thegreatbeyond/fish%20fmri.jpg” width=“453” height=“252” hspace=10 align=right border=0 />

Well, not quite red herrings, but Atlantic salmon. Allow me to explain. Reported on the Neuroskeptic and Neurolaw blogs this week is a study that aimed to demonstrate the risks of finding false positives in brain scanning studies – correlations that aren’t really there.

A group of scientists led by Craig Bennett at the University of California at Santa Barbara conducted their study with an unusual participant. From their Methods section:

“Subject: One mature Atlantic Salmon (Salmo salar) participated in the fMRI study. The salmon was approximately 18 inches long, weighed 3.8 lbs, and was not alive at the time of scanning.” Warning: below the fold it gets a bit fishy.


The salmon was presented with a series of photographs, and then asked to determine what emotion the individual in each picture was experiencing. The team then analysed tiny areas in the brain (voxels – like pixels but for volume) using basic methods for controlling for error. Surprisingly, report the team, “several active voxels were discovered in a cluster located within the salmon’s brain cavity.” They presented their poster at a plaice appropriate for the technique but not the species: the 2009 Human Brain Mapping conference.

Of course, Bennett’s group don’t mean to suggest that a post-mortem salmon is capable of perspective-taking. Cod forbid. They are making a serious point about the dangers of not taking account of false positives. When you image the brain using fMRI, you’re basically asking whether there is activation in each of thousands of voxels. Because there are so many voxels (130,000 in a typical fMRI scan), “the probability of a false positive is almost certain”, writes Bennett in the introduction.

Even so, some investigators don’t do enough to rule these false positives out, say the team. This follows in the wake (hake?) of a couple of papers earlier this year claiming that fMRI studies are subject to dodgy data analysis and spurious correlations. Mor(ay)over, there are options available in many fMRI stats packages for controlling the false discovery rate, they say. Relying sole-ly on standard statistical thresholds might leave you floundering.

Comments

  1. Report this comment

    Bob O'H said:

    I demand that fMRI techniques get a fair herring! Otherwise doctors will be skating on thin ice.

Comments are closed.