Example Creation in violation to the principle of onset asynchrony

[Asgard]

Dear all,

My questions here are in fact conceptural ones but they have been
bugging me for quite some time and I would really appreciate if
someone could provide me with some answers.

As the section "3.1. 1. Patterns" in the manual describes, there are
two types of patterns, one of which takes the form of raw voxel values
while the other uses voxel-wise beta weights.

In regard to the 1st type, I understand it is natural to extract the
pattern from individual scan volumes or the average of all the scans
within a single trial because it directly reflects the brain activity
level in response to the presented stimulus.

I would assume this requires that the start of each stimulus must be
temporally aligned to the onset of a scan, i.e.

____|(Stimulus Onset ___...

____|Scan1||Scan2||Scan3||Scan4|...

Otherwise, the correspondence between signal intensity and presented
stimulus is contaminated,

Trial time series: _______|(Stimulus Onset ___...

Scan time series: ____|Scan1||Scan2||Scan3||Scan4|... (Scan1 does not
purely reflect the brain activity in response to the stimulus)

Moreover, it is unreasonable to "shift the fMRI time series to account
for the hemodynamic delay" as in Kamitani & Tong and Haynes & Rees Nat
Neurosci published in 2005.

However, it seems to me that aligning the onset of scans to the onset
of stimulus poses a serious violation to a fundamental fMRI experiment
design principle known as Onset Asynchrony, as described in Human
Brain Function 2nd Edition by Dr. Rik Henson. Did I understand it
correctly? If so, how do we address this problem?


Many thanks and best regards
Ce

[J.A. Etzel]

I don't think that firm "answers" exist, but I can give a few opinions
and the guidelines I usually follow. I usually describe these issues as
'how to do temporal compression'.

Your first case (stimulus onset time-locked to image acquisition) is the
easiest. In this case I generally guess which images (Scan5, Scan6, etc)
should correspond to peak HRF and average those. This is straightforward
if the TR is short compared to the time period you want to temporally
compress (e.g. a twenty-second event and two-second TR) but can get
quite dodgy if the events and TR are close in time (e.g. events that
last a second). In these cases I generally think of analyzing single
timepoints or generating PEIs.

If stimulus onset is jittered in relation to image acquisition (your
second case) I follow a similar logic: if the jitter is minimal compared
to the TR (e.g. events start either half or three-quarters of the way
through a 1.5 second TR) or to the number of volumes being averaged
(e.g. a block design and 12 volumes are being averaged each time) I'll
probably just ignore the jitter. But if the jitter is large (e.g. 4 sec
TR and completely randomized stimulus onset) I'll think of PEIs again.

By PEIs I mean "parameter estimate images" - fitting a linear model
assuming the standard HRF and doing MVPA with the beta weights. To self
reference, I described some of this and presented a comparison of doing
averaging and PEIs on the same datasets in:
http://dx.doi.org/10.1016/j.neuroimage.2010.08.050 "The impact of
certain methodological choices on multivariate analysis of fMRI data
with support vector machines".

As a general strategy I look at the TR, stimulus timing, and event
duration for each particular experiment and question then think about in
which volumes the BOLD response we're looking for probably falls. If
it's a clear answer, I pick those volumes. If not, I design PEIs or
reformulate the question. None of this is a substitute for proper
experimental design and randomization, of course!

Jo

(also posted to
http://mvpa.blogspot.com/2012/06/temporal-compression-for-different.html )

On 6/4/2012 3:29 AM, Asgard wrote:
> Dear all,
>
> My questions here are in fact conceptural ones but they have been
> bugging me for quite some time and I would really appreciate if
> someone could provide me with some answers.
>
> As the section "3.1. 1. Patterns" in the manual describes, there are
> two types of patterns, one of which takes the form of raw voxel values
> while the other uses voxel-wise beta weights.
>
> In regard to the 1st type, I understand it is natural to extract the
> pattern from individual scan volumes or the average of all the scans
> within a single trial because it directly reflects the brain activity
> level in response to the presented stimulus.
>
> I would assume this requires that the start of each stimulus must be
> temporally aligned to the onset of a scan, i.e.
>
> ____|(Stimulus Onset ___...
>
> ____|Scan1||Scan2||Scan3||Scan4|...
>
> Otherwise, the correspondence between signal intensity and presented
> stimulus is contaminated,
>
> Trial time series: _______|(Stimulus Onset ___...
>
> Scan time series: ____|Scan1||Scan2||Scan3||Scan4|... (Scan1 does not
> purely reflect the brain activity in response to the stimulus)
>
> Moreover, it is unreasonable to "shift the fMRI time series to account

> for the hemodynamic delay" as in Kamitani& Tong and Haynes& Rees Nat

> Neurosci published in 2005.
>
> However, it seems to me that aligning the onset of scans to the onset
> of stimulus poses a serious violation to a fundamental fMRI experiment
> design principle known as Onset Asynchrony, as described in Human
> Brain Function 2nd Edition by Dr. Rik Henson. Did I understand it
> correctly? If so, how do we address this problem?
>
>
> Many thanks and best regards
> Ce
>

--
Joset A. Etzel, Ph.D.
Research Analyst
Cognitive Control & Psychopathology Lab
Washington University in St. Louis
http://mvpa.blogspot.com/

[Yuan Tao]

Hi Jo!

Just checked out that paper you post, excellent work, very helpful!
But there is one thing.. in the average case, you perform detrending
before averaging. Actually I did the other way around, averaging at
the last step.. You think is there any obvious wrong in what I did..?
Thanks :)

Yuan

> --
> You received this message because you are subscribed to the Google Groups
> "Princeton MVPA Toolbox for Matlab" group.
> To post to this group, send email to mvpa-t...@googlegroups.com.
> To unsubscribe from this group, send email to
> mvpa-toolbox...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/mvpa-toolbox?hl=en.
>

[J.A. Etzel]

Glad it helped! There are so many different legitimate ways to handle
these issues; it's absolutely critical to be explicit in methods papers.

It seems to me that either ordering could be defended in particular
cases. I usually do detrending before averaging since I think of the
detrending as part of the 'usual' fMRI preprocessing pipelines: drifts
often occur, and fMRI software has routines to take this out (e.g. as
part of fitting the linear models). But for the comparison paper it
seemed better to do the detrending after averaging, to better match the
procedure on the PEIs (and to simplify the procedure). I don't recall
comparing the voxel values from detrend-first vs. detrend-last; it'd be
interesting to do (or find a description from someone else).

I suppose, like most things, the best procedure will vary with dataset.
Sometimes plotting voxel timecourses before and after detrending or
averaging helps to choose the procedure: do the voxel values have an
obvious trend? Do the values change a lot between runs? Does it look
like the averaging captured the timecourse appropriately? You can't
check all voxels of course, but checking the impact in a few at random
is sometimes really helpful.

Jo

On 6/9/2012 5:39 AM, Yuan Tao wrote:
> Hi Jo!
>
> Just checked out that paper you post, excellent work, very helpful!
> But there is one thing.. in the average case, you perform detrending
> before averaging. Actually I did the other way around, averaging at
> the last step.. You think is there any obvious wrong in what I did..?
> Thanks :)
>
> Yuan
>