Supplementary Figure 1: Graphical depiction of the analysis pipeline.
From: Layer-dependent activity in human prefrontal cortex during working memory
(A) Multiple runs are collected per session with two different types (‘alpha / rem’, or manipulation versus maintenance, and ‘act / non-act’, or action versus non-action). Each run consists of interleaved images with (VASO) and without (BOLD) blood nulling. (B) Nulled and not-nulled images are motion corrected separately for the whole session. It is manually checked that the motion traces of nulled and not-nulled images are matching. (C) The time series are sorted by imaging contrast. (D) Runs of the same type are averaged within contrast. (E) Time series from all trials are averaged based on their task condition. In the 2 × 2 design used here, this results in 4 average trial types. Nulled images are corrected for BOLD contamination with a time-wise division of not-nulled images to provide a clean VASO contrast. (F) Based on the approximate location of activation in the low-resolution functional localizer, two layer ROIs are manually drawn on the T1-EPI anatomical images. (G) Time series for all four task conditions are extracted from the ROI of superficial and deeper layers for BOLD and VASO contrasts, resulting in 8 time series for each contrast, or 16 total time series.
