Hi all,
I am currently analysing data from an fMRI study, with a 2x2 between-within design. Patients vs controls reflect the between factor, and congruent and incongruent trials reflect the within factor (this is a Stroop task).
After seeing an interaction effect between the two factors in several brain clusters, I wanted to extract the beta-weights from these clusters to explore the underlying effects.
I used Marsbar to extract the raw beta-weights from congruent and incongruent trials, separately (so no contrast between conditions performed), and plotted the results for my two groups.
Many of the resulting beta-weights are negative, and I wonder what this means? Does it reflect a sort of 'deactivation'?
Any input will be highly appreciated!
Do you mean that you're getting negative beta values for the contrast between patients and controls? Your interpretation of that depends on whether you've coded "patients" or "controls" as the higher-valued dummy variable.
If, otoh, you're seeing negative beta values for the contrast between task and some kind of reasonable control (and you coded "task" as active and "control" as control correctly), then negative values would indeed indicate a decrease in activation in those areas.
Abigail,
Thank you for your response! See attached image for a figure that will hopefully clarify. The figure shows the beta-weights for a cluster, for healthy controls (HC) and a patient group (RAN), during two separate experimental conditions. There is a between-group difference in beta-weights in the congruent condition. My question relates to the interpretation of this effect, as controls show positive beta-weights, but my patient group show negative. What exactly does it mean when beta-weights are negative; how do you interpret them? I guess it relates to what a zero value means in this context.
Hi Lasse,
The beta is telling you the effect size and direction, presumably of MRI signal in your ROI. Whenever I'm doing fMRI-ROI analyses, I find myself re-reading parts of this paper:
http://scan.oxfordjournals.org/content/2/1/67.full.pdf+html
Hope this helps,
Reese
Most likely, in your case, when you run a general linear model (GLM) analysis, you are attempting to predict the voxel signal using a vector of onsets convolved with some hemodynamic response function. It's fundamentally the same as when you are carrying out any other linear regression: You are predicting some variable (Y) as a function of a predictor variable (X): Y = B(X)+E -- Y is a function of X, plus some variance, (E). B is simply the coeffcient of the predictor variable that leads to the best estimate of Y. In the case of fMRI, X is the idealized time course for the onsets of a particular condition. When B is high for a voxel, that means that the activation for that voxel closely follows the idealized time course for that condition. If B is near 0, that means there is little to no relationship between voxel activity and the idealized time course. B can also be negative, meaning that it is inversely related (e.g, whenever that particular condition occurs, activation actually decreases).
I think in general comparison of conditions (WS-factor) is the basis for interpreting your results. In your case, you found a clear increase of the BOLD response for congruent vs incongruent events in HCs, and the opposite pattern for patients.
I don't know which region you were looking at but I would have expected increases for incongruent events in e.g. ACC or frontal areas (in HC)
Thank you Reese for the article!
Chris, this is what I needed to confirm, that a negative beta weight actually relates to a decrease in activation, in relation to my onset vectors. Thanks!
Thanks Daniel,
There was an increase in frontal areas during incongruent trials, relative to congruent trials, for both controls and patients. These effects were located in the dorsomedial prefrontal cortex, and inferior / middle frontal gyrus (similar to previous studies using the same paradigm).
The figure included in an earlier post, however, were beta-weights from the amygdala. Both the left and right amygdala exhibited an interaction effect (condition x group), which is mainly due to the differential amygdala activation across groups in response to congruent trials, but not incongruent trials.
I always do a VOI analysis of a contrast-activated voxel, to characterize what is driving the A-B contrast. Is it that there is activation during event A or deactivation during B, or some combination of the two? Because I use event-related designs with trials every several seconds, I can use AFNI with mapfiles that enable trial-averaged dumpouts of the whole hemodynamic response averaged across all trials of that type (c.f. PMID 18672069 ). That gives a graphical answer right there. IMO one shouldn't just plug and chug contrasts using turnkey software without knowing what's going on in the trial-type-averaged BOLD signal volume by volume.
- Badges
- Science topic
- Similar topics
- Correction