Project 2: Discussion
Signficant Network-level emotion processing activity observed
All three approaches contained boundraries that demonstrated significant ‘emotion processing’ activity, including the left and right fusiform as well as the amygdala of the Meta analytic ROIs, the Visual-Central network (VIS-C), and dorsal attention network B (dATN-B) of the individualized networks, and the Visual-Central network (VisCent) of the group networks.
The left and right fusiform regions are involved in face processing.
The Visual-Central (VisCent & VIS-C) networks are involved in general visual processing.
‘Emotion processing’ activity was greatest within the right fusiform and the Individualized Visual-Central Network (VIS-C).
Significant activity in the Dorsal Attention Network B (dATN-B) of the individualized networks suggests greater general attentional demands during the faces condition compared to the shapes condition.
Dorsal Attention A (DorsAttnA – the group network that is topologically equivalent to dATN-B) exhibited significant positive activity in both the faces and shapes conditions, but the difference in activity between the conditions was negligible. Suggests Group networks are not sensitive enough to capture nuanced differences in activity, while individualized networks can capture subtle differences. Although the Visual-Central network (VisCent) in the group networks captured significant ‘emotion processing’ activity, the degree of activation was notably reduced compared to the equivalent individualized network (VIS-C).
Confirmation that activity is actually occuring at the level of the network
network-wide increases in vertex activity for networks, suggesting significant network-level engagement.
Quantitative analysis of vertex activity supported the presence of network-level structure, with the visual-central network and dorsal attention network B showing the highest percent of ‘emotion processing’ activity above zero.
Network-level differentiation (moderate ICCs) further corroborates the visual evidence, indicating meaningful network-level activity in the data.
Emotion processing activity was well outlines by network boundaries. Boundary-activity overlap is particularly striking for higher order networks such as control network A and default network A. Other work examining extent of individualized network-activity overlap were directly attempting to engage particular network/region function (Braga et al., 2020; Chong et al., 2017; DiNicola et al., 2020; Gordon et al., 2017). In contrast, the current task was not designed to engage network-level function, yet network boundaries still captured the observed activity well.
Additional networks showed notable network-level activity during a subset of sessions.
In sum, vertex activity shows meaningful network-level structure, suggesting that using network boundaries to dimensionally reduce the data is a valid approach to analyzing task activity data.
A network-level approach to analysis may provide a way to make sense of the activity observed in this task that is differentiated from a region of interest approach.
Impact of results:
Observed meaningful network-level activity in canonical emotion processing task not designed to elicit network-level function.
Suggests that network-level activity is a robust feature of the data, and that network boundaries can be used to examine data in a meaningful way.
future work – examine associations of activity with external factors to further validate this activity
[INSERT MORE DISCUSSION HERE]