Fig. 3: EMA temporal network architecture and symptom-related modulation of behavioral dynamics. | npj Digital Medicine

Fig. 3: EMA temporal network architecture and symptom-related modulation of behavioral dynamics.

From: Shedding light on the dynamic interplay of positive and negative symptoms of psychosis with Behavioral Tractography

Fig. 3

Figures were generated using dedicated open-source software (https://dev.mlnetwork-diplab.ch/; https://github.com/andreaimparato/Behavioral-Tractography-Toolbox/tree/main/MLNetwork). For each figure, a link to an online platform provides an interactive 3D visualization. 1A, 1C, 2A, 2C depict Multi-Layer Temporal Networks (MLTNs) constructed separately for four populations: High-Symptom-Intensity (1A), Low-Symptom-Intensity (1C), Predominantly-Positive-Symptom (2A), and Predominantly-Negative-Symptom (2C). The 3D multilayer structure represents behavioral dynamics. Cross-sectional connections appear within temporal layers, while longitudinal edges extend along the Z-axis, linking variables from Temporal Layer 1 (TL-1, left) to Temporal Layer 2 (TL-2, right). Node placement within each layer follows Network Dimensionality Reduction, grouping variables that are closely related cross-sectionally. The slope of longitudinal edges reflects the likelihood of dynamic interactions across temporal assessments. Edge thickness indicates the strength of associations from mixed-model regression, while arrows denote edges belonging to shortest paths connecting variables across layers. Edge transparency reflects the number of shortest paths traversing an edge. Node color encodes longitudinal betweenness centrality. A reverse-coded green-to-yellow gradient across TL-1 and TL-2 highlights temporal flow: TL-1 variables shade from green to yellow according to their propensity to act as gateways toward future states, while TL-2 variables shade from yellow to green to represent their role as funnels from past states. Variables acting as longitudinal hubs are highlighted in bold. Node size indicates summed connectivity strength. 1B, 2B show MLTNs modulated by SIPS dimensions. Edges are color-coded to indicate effects of SIPS dimensions on association strength between EMA variables, while node color reflects their influence on average intensity. 1B represents SIPS-Dimension 1 (overall symptom intensity): red edges and nodes denote stronger associations or higher intensity in High-Symptom-Intensity, while blue indicates stronger associations or higher intensity in Low-Symptom-Intensity. 2B represents SIPS-Dimension 2 (Predominantly-Positive vs. Predominantly-Negative Symptoms): red denotes stronger associations or higher intensity in Predominantly-Positive Symptoms, and blue in Predominantly-Negative Symptoms. Links to 3D interactive visualization: https://dev.mlnetwork-diplab.ch/3dvisualizer/net_dim1_hsi/. https://dev.mlnetwork-diplab.ch/3dvisualizer/net_dim1_lsi/. https://dev.mlnetwork-diplab.ch/3dvisualizer/net_dim2_pns/. https://dev.mlnetwork-diplab.ch/3dvisualizer/net2_dim2_pps/. https://dev.mlnetwork-diplab.ch/3dvisualizer/dim2_pns_pps/. https://dev.mlnetwork-diplab.ch/3dvisualizer/dim1_lsi_hsi/.

Back to article page