Reproducibility of Activation Map across Sessions and Groups
Table 1 describes the spatial overlap across sessions and
groups using dice coefficients. In general, the reproducibility of
group-level maps was moderate. The alerting network had the highest dice
coefficient across sessions (HC: 0.59, PD:0.46) and groups (Session 1:
0.62, Session 2: 0.49). The group-level reproducibility for the
executive network was the worst, except that the PD group’s spatial
overlap across sessions was about half (dice = 0.46). The spatial
overlap across groups for the executive network was unsatisfactory,
especially for the second session (dice = 0.15). In the
supplementary material (Table S5), we also provide results for
dice coefficients using varying thresholds, which indicated a similar
finding. In addition, we conducted a region-based analysis using the
Schaefer-Yeo atlas to help readers better grasp the reproducibility of
attention networks across brain regions. As shown in Figure 2 ,
the reproducibility at the region level was poor. The region-based
visualization provided a coarse summary of the attention networks across
groups and sessions. However, it should be noted thatFigure 2 is for illustration purpose as averaging activations
at the region level might be too coarse to reveal the real spatial
activations.
Table 1. Dice
Coefficients of the Three Attention Networks across Sessions and Groups