Are All FMRI Results Wrong?
Vložit
- čas přidán 22. 07. 2024
- Overview of the Eklund et al (2016) paper on false positives in FMRI. Click "Show More" for details about false positive rates, and a tutorial on how to address them.
Anders Eklund and colleagues published a paper last summer showing that cluster correction - one method that FMRI researchers use to test whether their results are statistically significant or not - can lead to high false positive rates, or saying that a result is real, when actually it is a random occurrence that looks like a real result.
Their calculations showed that about 10% of FMRI studies are affected by this error (tinyurl.com/jaomsgs). However, keep in mind that even if a study is at risk for reporting a false positive, doesn't mean that their result is necessarily spurious. As with all results, one must go to the original study and take into account the rigor of the experimental design and whether the result looks legitimate.
These flaws have been addressed in recent versions of AFNI, an FMRI software package. The steps to use these updated programs can be found on the blog here: tinyurl.com/j5vafsb - Jak na to + styl
Most parsimonious and objective commentary on the subject I've seen/heard thus far, bravo dude!
It's only six months late, but thank you!
Thanks, Andy. That was helpful. This was an extremely frustrating cycle of reporting at the time because of the grossly inappropriate report of 70% error rate in fMRI that made headlines. One of the principal caveats to Anders' criticism was that it implied that fMRI researchers simply take what they are given when they put EPI data through the analysis mill. That may have been the case in 1992 when the technology was new and models of activity were raw, but over the years, this has changed. I would never accept a manuscript submission that looked at functional activity of "the brain during a task". Yet, this is what the article implies: spaghetti hit the wall and stuck, so off to Nature it is sent.
Just a rambling.
Just wanted to thank you. I just started my PhD Thesis and have to analyze fmri data the first time alone. You help me alot!
Glad to hear it helps!
Good Presentation!
I always wondered that with the extend thresholding, we are disregarding activated voxels that are smaller than the extent threshold. However, considering a voxel alone can contain thousands of neurons, and a neuron cluster of interest could reside in just one voxel or several voxels, would not it be too harsh to use extent based thresholding since it will only look for mass activations instead of specific ones? For example, if one would like to study mirror neurons it would be impossible for him to use fMRI unless thousands of them are activated at the same time, and yet there are still articles using fMRI for that purpose.
Cluster thresholding isn't the only correction option; you can also do voxel-based threshold, such as Bonferroni or False Discovery Rate. Bonferroni is considered much too stringent, however, since virtually all voxels share some variance with their neighbors, and therefore should not be treated as independent tests.
I don't know too much about mirror neurons, but would it be a stretch for a large group of them to be active at once to represent a certain gesture or action? Much of this depends on your voxel resolution, and how widespread you believe the mirror neurons to be (although I've heard that they are up for debate).
Hi Andrew, Is there any chance you could have my son's 3t Mri reviewed. Two neuroradiologist have different opinions regarding FCD. Happy to pay whatever it costs.
Hi Salmir,
I'm not a clinician, so I can't make a formal diagnosis; I would help you if I could.
Regards,
-Andy
hi Andrew..thank you ....i wanna do my Ph.D dissertation by FMRI in IRAN...you think it is possible to consult with you if it is necessary .
Hi Arad,
Please send me an email at ajahn@umich.edu if you would like to talk more about this. Thanks!
-Andy
Please help me understand the meaning of non independence error in FMRI
Hi Chileya,
I'm working on a video about that right now, and I will publish it this weekend.
Best,
-Andy
Hello Dear Andrew. Is there any chance you could make a few tutorials on group analysis such as t test and paired t test? I would highly appreciate that!
Hey Alaleh, I believe there are a couple of tutorials already up about doing t-tests in SPM (search for SPM Model Specification and Estimation, and SPM Contrasts). I've been meaning to update those with better animations, but those should help you in the meantime.
Best,
-Andy
I will check them out. Thanks, your videos are a lot of help!