This is a technical discussion, intended to be helpful to researchers designing or analyzing fMRI studies of bladder filling that use a repetitive infuse-withdraw paradigm. Because the fMRI signal may not be particularly stable over long periods, the signal is usually high-pass filtered at a period of about 128 s. (A cut-off of 84 s was used in this study, for mathematical reasons.) At a given voxel, it is then necessary to measure the difference between the signal of interest and a local baseline value a few seconds earlier or later (a "contrast"). Because the fMRI signal is quite noisy, it is helpful to average repeated measurements made over a time of perhaps a few minutes.
The protocol
used in the current study is shown schematically on the left.
There are 4-6 blocks blocks of measurement, 2 with only small
volume in the bladder, and 2 following filling of the bladder
to a strong desire to void, plus optional 5th and 6th blocks if
the subject can tolerate them.
Each block consists of 4 repetitions of a basic pattern - pause / infuse 22 ml / pause / withdraw 15 ml. Thus the volume in the bladder tends continually to increase, avoiding accommodation and in fact ensuring that by the end of scanning there is a very strong desire to void. This has some advantages, but also some disadvantages if the subject shows actual detrusor overactivity during scanning, as the desire to void can become so intense that scanning has to be stopped prematurely.
Each of the phases of the basic pattern (pause, infuse,withdraw, infuse) lasts 10.5 s, equivalent to 7 whole-brain scans because the scan repetition time TR is 1.5 s. In the analysis, each phase is treated as a unit, and the fMRI signal is obtained from a suitable contrast, e.g. infuse - withdraw. This design assumes that the signal is essentially constant during each phase of the basic pattern, or at least that the average signal during each phase is representative of that phase.
One desirable condition is that there should be several scans during each phase so that an average can be taken. In our case there are 7 scans per phase, which seems adequate. However, there are other considerations. fMRI measures the hemodynamic response to neuronal activation, which is delayed and, more importantly, smoothed over about 5 s. Thus if the phases are to be recognizable in the fMRI signal they should last noticeably longer than 5 s. Finally, there are temporal correlations between successive scans; again the duration of each phase should be considerably longer than the correlation time. If all these conditions are met one would expect the 4 phases to be recognizable in the fMRI signal, and the use of the average value in each phase to calculate the desired contrasts to be acceptable.
The signal at an activated voxel can be inspected in SPM by plotting an F-contrast (showing each of the 4 phases) against time. An example is shown in the figure below.
The
fitted curve shows the 4 phases but there is considerable smoothing,
so that they are not as distinct as one might like. Clearly the
duration of the phases is only just long enough to allow successive
phases to be distinguished.
In
order to determine what contrast should be used to calculate the
fMRI signal, a plot of the mean and CI of the signal in the 4
phases is helpful (see figure on left). In the study reported
by us in J Urol, no subject, even among those with detrusor overactivity
on prior investigation, showed actual detrusor overactivity during
scanning. In general, as shown in the figure (left) the greatest
fMRI signal was registered during infusion and the smallest or
near-smallest during withdrawal. The signal values during the
2 pause conditions were generally intermediate and rather variable
from one block to another. Actual subject experience indicates
that bladder sensation is greatest during infusion and smallest
(perhaps zero) during withdrawal. Thus withdrawal seems to represent
a baseline corresponding to minimum or zero afferent input to
the CNS, and it is natural to take (infusion - withdrawal) as
the basic contrast representing signal strength.
Should the above conditions
not be satisfied then a different contrast might be more useful.
For example, if the scan repetition time is 3 s, and the duration
of the phases is only 5 s (an actual example that I have seen
in another laboratory), then the effect of smoothing is to blur
and obscure the boundaries between the 4 phases. The fitted fMRI
signal then shows a quasi-sinusoidal oscillation that may peak
during the pause following infusion (pause1), or reach its minimum
during the pause following withdrawal (pause2). In this case a
different contrast might be appropriate, for example pause1 -
pause2, or perhaps (infusion + pause1) - (withdrawal + pause2).
Some experimental plotting of signals and means would be necessary
to come to a conclusion.
We are performing a new series of measurements, using the same protocol, in older women. Four subjects have showed pronounced detrusor overactivity during actual scanning. This seems to lead to a different dependence of signal on infusion and withdrawal. Probably the contrast (infusion - withdrawal) is no longer the most appropriate. At any rate it is clear that thought and discussion about the protocol and the best contrast to use under various conditions are needed. It is difficult to discuss these things in formal papers, and so if anyone cares to email me about their thoughts or experiences I will put them on this web page.