Whale of a story

Whale of a story

No Red Herring here. Bennett et al should be applauded for their attempt to publish negative results or more accurately a false positive. I’m curious if the experiment described in the Wired.com article was not published due to the fact that it isn’t news. I would be surprised if anyone doing functional imaging is unaware that incorrect thresholding of the images/data could lead to a false positive.

In case you are interested, I’ve discussed the differences between anatomic and functional images here: http://goo.gl/kxl8b and here: http://goo.gl/6apzH

They were able to publish their thoughts on the need for careful statistical analysis of fMRI, albeit with no dead salmon images.

http://scan.oxfordjournals.org/content/4/4/417.long

Thanks for sharing Jennifer Ouellette

#ScienceSunday

#fMRIblob  

Originally shared by Jennifer Ouellette

And here’s a 2009 post on the classic fMRI scan of a dead salmon that showed a bit of brain activity. Supposedly. 🙂

http://www.wired.com/wiredscience/2009/09/fmrisalmon/

13 thoughts on “Whale of a story”

  1. Hehe. Were I the reviewer of this study, I would have asked for more salmon to be subjected to the fMRI. At least 6 each of farm raised and wild, as a minimum. Also, why didn’t they show the subjects images of other salmon? As control, they should have imaged a dead human and monitored their response to images of salmon. Then subtracted the voxels to produce a differential image. Both before and after dinner. Also, why was the t-value pseudocolored in red? I believe that’s the real red herring.

  2. Good morning Rajini Rao If it was freshly caught salmon, I could see where the referee might be miffed enough by them not sending a sample along with the manuscript (as supplemental data).

  3. Good morning Feisal Kamil it’s not too hot yet. We escaped the heat yesterday. Would have been nice if it rained.

  4. Gnotic Pasta you adjust the SNR by adjusting the imaging or post-processing parameters. In the articles they wrote, they emphasize careful use of statistics to “filter” out erroneous results. Another straightforward way is to balance spatial resolution with SNR, i.e., make the image a little less detailed for higher SNR.

Leave a Reply