Dynamic network topological properties for classifying primary dysmenorrhoea in the pain-free phase

Primary dysmenorrhoea (PDM) is known to alter brain static functional activity. This study aimed to explore the dynamic topological properties (DTP) of dynamic brain functional network in women with PDM in the pain-free phase and their performance in distinguishing PDM in the pain-free phase from healthy controls.

Thirty-five women with PDM and 38 healthy women without PDM were included. A dynamic brain functional network was constructed using the slide-window approach. The stability (TP-Stab) and variability (TP-Var) of the DTP of the dynamic functional network were computed using the graph-theory method. A support vector machine (SVM) was used to evaluate the performance of DTP in identifying PDM in the pain-free phase.

Compared with healthy controls, women with PDM had not only lower TP-Stab in global DTP, which included cluster clustering coefficient (C_p ), characteristic path length (L_p ), global efficiency (E_g ) and local efficiency (E_loc ), but also lower TP-Stab and higher TP-Var in nodal DTP (nodal efficiency, E_nod ), mainly in the prefrontal cortex, anterior cingulate cortex, parahippocampal regions and insula. The TP-Stab and TP-Var were significantly correlated with psychological variables, that is positive emotions, sense of control and meaningful existence. SVM analysis showed that the DTP could identify PDM in the pain-free phase from healthy controls with an accuracy of 79.31%, sensitivity of 82.61% and specificity of 76%.

Women with PDM in the pain-free phase have altered global DTP and nodal DTP, mainly involving pain-related neurocircuits. The highly variable brain network is helpful for identifying PDM in the pain-free phase.

This study shows that women with primary dysmenorrhoea (PDM) have decreased stability of dynamic network topological properties (DTP) and increased DTP variability in the pain-free phase. The altered DTP can be used to identify PDM in the pain-free phase. These findings demonstrate the presence of unstable characteristics in the whole network and disrupted pain-related neurocircuits, which might be used as potential classifiers for PDM in the pain-free phase. This study improves our knowledge of the brain mechanisms underlying PDM.