Fig. 4: Network-based diffusion model for predicting CT maturation in Statistical Model I. | Nature Communications

Fig. 4: Network-based diffusion model for predicting CT maturation in Statistical Model I.

From: Structural connectome architecture shapes the maturation of cortical morphology from childhood to adolescence

Fig. 4

a Schematic diagram of nodal diffusion processes through multiscale WM edge paths. The orange color represents the edges and nodes at the mth neighboring scale of a given node i (red). The diffusive profiles of all nodes form the diffusive probability matrix at each neighboring scale. b The curve of the average diffusive probability of whole-brain nodes within or between cortical systems. It illustrates that the within-system diffusion probability was greater than 0.5 at the first scale and then decreased along neighboring scales. c Significant Pearson correlations between the predicted CT maturation and the observed CT maturation (t-value, 1000-node resolution) against 1000 rewired tests (light blue boxes) and 1000 spin tests (deep blue boxes). Boxes represent the IQR with the lower and upper boundaries correspond to the 25th and 75th percentiles, and an inside line indicating the median. The whiskers represent values within 1.5×IQR and individual points represent outliers. Asterisks denote a significance level at p < = 0.001, one-sided. The scatter plot depicts the Pearson correlation between actual and predicted CT maturation (r = 0.75, p = 5.11 × 10−178, two-sided, linear fit (central line in red) with a 95% confidence interval (shadows in gray)) at the 3rd neighboring scale, which exhibited the highest accuracy. d Schematic of the diffusion-based approach used to identify dominant regions. e Regional distributions of dominant likelihood (cosine similarity between nodal diffusion profiles and CT maturation map (t value)) at neighboring scales of 1−9 (top panels) and the spatial distributions of dominant regions (pspin < 0.05, one-sided, bottom panels). f The conjunction map of dominant nodes shows the probability of each node being identified as a dominant node across neighboring scales. g The diffusive probability maps of two representative dominant nodes separately in the prefrontal (PFC, top panels) and inferior parietal cortex (IPC, bottom panels) at each neighboring scale. Brighter color represents a greater diffusive probability. The right panels show the diffusive probability within or between systems. As the neighborhood scale expands, the diffusion paths of two nodes spreads from local to distributed communities.

Back to article page