Development of top-down cortical propagations in youth

The hierarchical organization of the cortex affords bottom-up sensory integration and top-down control. Notably, top-down hierarchical processing necessarily involves activity propagating through space from higher- to lower-order areas. However, most fMRI studies of hierarchical processing have chiefly quantified activity fluctuations in fixed regions over time rather than accounting for activity propagations over space. This gap is critical for adolescent neurodevelopment: the emergence of mood disorders and mortality via risk-taking behaviors and are in part attributed to impairments top-down control, but it is unknown whether top-down propagations become more prominent with development in youth. Here, we fill these gaps by adapting an established computer vision method—optical flow—to quantify the direction of infraslow (<.1Hz) cortical activity propagations in a large sample of youth. Optical flow effectively recovered activity directionality in simulations, uncovered significant enrichment for bottom-up and top-down propagations in every participant tested, and revealed that top-down propagations increase with task-demands and over neurodevelopment.
Image preprocessing followed an updated protocol from the Human Connectome Project MRI pipeline including gradient, bias field, and EPI-distortion correction, template normalization, re-alignment, boundary-based registration, and gaussian smoothing. Functional images also underwent bandpass filtering (0.008-0.09 Hz) and 27-parameter nuisance regression. Our samples were the Human Connectome-Development Project (HCP-D, n = 388) and the Midnight Scan Club (MSC, n = 10).
We adapted spherical optical flow to spherical registrations of each participant’s preprocessed functional timeseries. At its core, this technique estimates a regularized vector field that explains the displacement of signal between two temporally adjacent frames. Only temporally adjacent frames with low motion that were in the same low-motion continuous (>10 frame) segments were subjected to optical flow analyses.
To operationalize optical flow vectors as either bottom-up or top-down, we extracted the gradient vector field (∇) of a scalar map of functional hierarchy (FH) for a vector field describing hierarchical ascent (bottom-up directionality). Activity flow moving within 90 degrees of ∇FH was considered bottom-up, and activity flow >90 degrees from ∇FH was considered top-down. Head motion was covaried for in all results.
Despite noise and spherical inflation of the simulated data, measurement of the vector fields obtained from the simulated data revealed a prominent alignment of the directions recovered by optical flow with the ground truth relative to conservative spatial permutations (p< 0.001).
In applying this approach to real data from HCP-D, we found a predominance of both bottom-up and top-down hierarchical propagations, which formed a bimodal distribution. These bimodal distributions were evident at the single participant level: angular distributions of propagations were bimodal in every participant in the HCP-D sample (p < 0.001 for all participants) and every participant in the MSC sample (p < 0.013 for all participants) relative to conservative spatial permutations.
Having described hierarchical activity propagations at rest, we next sought to determine whether they were modulated by the performance of a demanding cognitive task. We compared propagations observed during rest with those present during a modified Go/NoGo task, where top-down control is required to suppress reflexive button-pressing. Analyses revealed more top-down propagations during task than rest (n = 281, t = 2.37–13.97, p < 0.05).
Given increases in top-down cortical propagations during a cognitive control task and that cognitive control improves with age, we next evaluated whether top-down propagations increased with age. Maturation from 8-22 years old was associated with widespread increases in the proportion of top-down propagations across the cortex (Δ Adjusted R2 = 0.01–0.19, p < 0.05). This increased prevalence of top-down propagations was evident even when measurements were averaged across the entire cortex (Δ Adjusted R2 = 0.14, p = 1.7 × 10−14). When covarying for the most-reported finding in functional cortical development, default mode segregation, the relationship between age and increased top-down activity movement was unimpacted and yielded an effect size several times larger than that observed for default mode segregation.
Analysis of the movement of cortical activity over space revealed substantial enrichment for bottom-up and top-down propagations in every participant. Top-down propagations increased during the performance of a demanding cognitive task and with maturation. These results implicate cortical propagations as a novel mechanism of hierarchical processing and neurodevelopment.