Dear Mike,
I agree with Karl that around 24 replications would be sufficent if the
signal is sufficently high relative to noise and you are looking for
large robust differences, e.g. betwen stimulation and baseline states.
We have also used this number of conditions in a parametric study of
tone intensity. However, the most appropriate number of repeats per
condition depends upon many factors including the protocol, the SNR,
the nature of the response elicited by the stimuli and the nature of the
differences that you are seeking to detect.
Our original use of the term "sparse imaging" refers to a protocol in
which a stimulus condition contained a continuous auditory signal for
between 10-14 s alternated with a silent baseline condition (see HBM 99
vol7(3) pp213-223). In our study, using the sparse imaging method, a
single volume of images was acquired at the transitions between each
epoch. Efficiency was gained since the haemodynamic response to the
stimulus had effectively reached a plateau 10-14 s after stimulus onset
thus maximising any differences between ON/OFF states. However this
method tells you nothing about the time course of the response - for
that you would need to use a sparse imaging protocol as used by Belin et
al (99 neuroImage) for example.
I spoke to Ingrid Johnsrude about this type of experimental design a
little last Friday and so you might like to discuss it with her. If you
are planning to measure differences BETWEEN listening states then you
may need to pilot your study to optimse efficiency as this might require
greater sensitivity than simple ON/OFF contrasts.
Hope this helps.
Debbie Hall
MRC Institute of Hearing Research
Nottingham
***********************
Karl Friston wrote:
>
> Dear Mike,
>
> > We are intending to run a sparse imaging study but are unsure about how
> > many volumes are necessary per condition. The figure we have heard is
> > 25-30, but are unsure where this figure comes from and whether it is
> > relevent for our experiment.
> >
> > We intend to auditorily present 3 or 4 words per 12 second TR and
> > subjects will perform a non-word identification task at the end of the
> > stimulus presentation part of the TR. We will be using a 3T magnet.
>
> This is a difficult question because there is no definitive answer. By
> 'sparse imaging study' I assume you mean an event-related study with
> sparse events. As you probably know this is the least efficient design
> one has to contend with. Assuming that the task requirements enforce
> this design I would note that an inter-trial interval of slightly more
> than 12 seconds would be more efficient (ideally about 16 seconds
> assuming a fairly short TR). Our experince is that results are
> obtained with as few as 16 events (provided they elicit a reasonable
> response), but generally more (e.g. 32) is better. At the end of the
> day I think you just have to do the experiment and see if you get
> results.
>
> I hope this helps - Karl
|