Friday, July 18, 2008

Is the MTL One Big Happy Region?

A recent post in BRAINETHICS spawned an interesting discussion about the limits (and truthiness) of image coregistration. Thomas Zoëga Ramsøy spotted an egregious error in an fMRI study that appeared in a high-profile publication last year:
Big fMRI error in Science!!!

. . .

HOW CAN YOU GET A SCIENCE PUBLICATION WITH A HUMONGOUS ERROR?

Take a look at this image. It’s from a 2007 article in Science by Depue et al.

It’s supposed to show activation in the hippocampus and amygdala. Looks innocent, right? Let’s take a closer look.
Since this paper was the subject of The Neurocritic's entry I Forgot... from July 16, 2007, it was time for me to take a closer look as well, since I missed it the first time around.


from Fig. 2C (Depue et al., 2007). Functional activation of brain areas involved in memory processes and emotional components of memory. Blue indicates greater activity for Think trials than for No-Think trials. Conjunction analyses revealed that areas seen in blue are the culmination of increased activity for T trials above baseline as well as decreased activity of NT trials below baseline.

In brief, this figure is supposed to illustrate that the hippocampus showed less activation when subjects were instructed to prevent the memory of a previously presented image from reaching consciousness. The authors (Depue et al., 2007)
used the Think/No-Think paradigm (T/NT) of Anderson and colleagues, in which individuals attempt to elaborate a memory by repetitively thinking of it (T condition) or to suppress a memory by repetitively not letting it enter consciousness (NT condition).
However, the blue blobs are outside the hippocampus, which is labelled in yellow.

The critique in BRAINETHICS continued:
I spent quite a while figuring the figues out. Did the yellow names indicate the blobs or where the structures actually are? For one thing, the blue blobs don’t fit into amygdala or hippocampus, but rather the entorhinal cortex.

Another coronal slice from Fig. 2C (Depue et al., 2007).

But let me comment on two big errors related to this slice.

First, the hippocampus is not present on this slice, so why put the name there? And why put it that lateral? This is really bothering. Do the researchers (and reviewers) really think that the hippocampus has anything to do here?

Second, the rightmost activation blob is centered in white matter. Hmm.. would that not give you the opportunity to speculate whether your coregistration was correct? I would.

An anonymous commenter disagreed:

I think you’re getting all worked up over very little. The slices shown range from y=-22 to y=5, so the clusters of activation probably extends beyond the hippocampus to include some other regions. It is rather weird that they would choose to label something clearly outside of hippocampus as such. Entorhinal cortex is part of the hippocampal system, so there’s nothing wrong with the “hippocampus” label.

That's interesting...aren't there entire research programs devoted to differentiating the roles of the hippocampus and the entorhinal cortex in episodic memory (e.g., see Lipton & Eichenbaum, 2008 for starters)?

Anonymous continued:
It’s not as if fMRI is going to be credible at distinguishing the various parts, at least not with their design.
Then why label the blue blob in question "the hippocampus"?

Dr. Ramsøy also disagreed with Anonymous:
IMO, too many people make fMRI localization in a much too loosely or unconstrained way. What do they do with these kind of activations? Use the SPM Anatomy toolbox… which definitely is wrong in this region. Or just the use of spatial normalization is known to induce errors.

. . .

So you get the anatomy wrong. What you call the hippocampus may be the entorhinal or perirhinal, what you call the amygdala may the the temporopolar cortex. So what, right? Isn’t this just one big system? Just as Squire, Zola and the others have claimed a long time ago? Well, if you look at the literature, there is a big discussion (just as ardent as that of the fusiform face area) that is vigorously debating the role of different medial temporal lobe structures. That is why this is important. It DOES matter what you label your blobs as, and in particular whether you check your images properly. People are way too sloppy about this (still).
Too sloppy for Hippocampus, but not for Science...

Subscribe to Post Comments [Atom]

0 Comments:

Post a Comment

<< Home

eXTReMe Tracker