Figure 1.

Analysis pipeline linking multimodal data in IUD users who completed a mental rotations fMRI task and provided saliva for hormone assays. (A) A person-specific neural network generated by group iterative multiple model estimation (GIMME) for one IUD user. Black nodes show putative mental rotations network regions, and blue nodes show default mode network regions. Solid lines are contemporaneous (same-volume) connections, and dashed lines are lagged (next-volume) connections. Thick black lines are group-level connections significant for at least 75% of the sample, but estimated for all IUD users, and thin gray lines are individual-level connections unique to this IUD user; all participants had corresponding estimated networks (though not depicted here). All connections also have a direction (positive or negative) and beta weight associated with them (also not depicted here). This woman’s network fit her functional data well (χ² (112)=652.60, p<.001, RMSEA=.135, SRMR=.039, CFI=.955, NNFI=.924). R, right; L, left; IFG, inferior frontal gyrus; Par, parietal; LP, lateral parietal; sPar, superior parietal; MPFC, medial prefrontal cortex; PCC, posterior cingulate cortex. (B) Average neural network densities extracted from the person-specific GIMME networks of all IUD users (and divided by overall network complexity), with error bars showing standard deviations. (C) Correlations among multimodal data, including mental rotations task performance in the scanner, endogenous hormone levels, and neural network features, including overall complexity and network densities. Color-coded correlations are shown in the matrix, with dark red reflecting strong inverse relations through dark blue reflecting strong positive relations.