I would appreciate some clarification regarding the interpretation of source-level activity obtained with a dSPM/MNE pipeline. I am analyzing EEG data from a continuous tracing task with two conditions (Condition A vs Condition B). Source reconstruction was performed in Brainstorm using the Colin27 template, a three-shell spherical head model, and MNE with unconstrained dipole orientations followed by dSPM normalization.In source space, activity was z-scored relative to a pre-task baseline. The three unconstrained dipole components (x, y, z) were combined using the L2 norm.
For each participant, I averaged the resulting source amplitude over a prolonged task interval (25 s) and compared Condition A vs Condition B conditions using a paired permutation Wilcoxon signed-rank test.
My question concerns the direct interpretation of the resulting statistical current maps. When a region shows significantly higher broadband source amplitude in one condition (e.g., warm colors / positive T-values in the Condition A):
Can this be directly interpreted as greater cortical activation/engagement?
Or, because it is a very large time window, am I actually just seeing the high oscillatory amplitude of low frequencies (like alpha and beta)? If it is the latter, an amplitude increase in the alpha/beta bands functionally means synchronization (idling/gated cortex), which is the exact inverse of cortical activation.
Hello @marievavl, before addressing the questions, I'd like to point that this step looks problematic:
By doing so, the L2(z-score_x, z-score_y, z-score_z) will result in an always positive number, moreover the combined z-scores for each direction are computed with a different baseline_mean_{x|y|z} and baseline_std_{x|y|z}. Thus this aggregated z-score is hard to interpret by it self. Even when individual directions have negative z-scores, this information is lost.
A better approach, would be to compute L2 (in the baseline and task) before z-score, in that sense the z-score will represent strength of the source with respect to baseline.
Here the average is on the the always-positive z-scores (described above)
Not really, as the it is the comparison of the always-positive z-scores, which represent the amplitude difference (without sign) with respect to baseline. So, the interpretation of the sign of the t-values is just which condition was more different to baseline, but it is ambiguous if activity in A was higher or lower than baseline, same for B.
From the time average, it is not possible to know which band is contributing to the total activation.
Source activity follow the 1/f power distribution, so lower frequencies contribute more than higher to the total activation. To do this comparison in the freq domain is needed