Characterization of CNS Network Changes in Two Rodent Models of Chronic Pain

Abstract

Neuroimaging in rodents holds promise for advancing our understanding of the central nervous system (CNS) mechanisms that underlie chronic pain. Employing two established, but pathophysiologically distinct rodent models of chronic pain, the aim of the present study was to characterize chronic pain-related functional changes with resting-state functional magnetic resonance imaging (fMRI). In Experiment 1, we report findings from Lewis rats 3 weeks after Complete Freund’s adjuvant (CFA) injection into the knee joint (n = 16) compared with the controls (n = 14). In Experiment 2, Sprague–Dawley rats were scanned 2 weeks after partial sciatic nerve ligation (PSNL) (n = 25) or sham surgery (n = 19). CFA and PSNL induced typical behavioral patterns consistent with inflammatory and neuropathic pain, respectively. Functional magnetic resonance imaging analyses comprised (1) independent component analysis (ICA) decompositions, (2) assessment of graph measures, (3) seed-based functional connectivities, and (4) predictions of chronic pain based on supervised machine learning. In both models, we detected changes in default mode network (DMN) activity. Local and global graph measures were generally similar across groups. However, regardless of the pain model, we observed a significant reduction in the betweenness centrality hub disruption index (HDI) in chronic pain compared with the controls. Finally, employing supervised machine learning in combination with a deep learning approach, chronic pain became predictable based on the functional connectivity patterns. The results indicate changes in DMN activity and betweenness centrality HDI in chronic pain. The predictability of chronic pain using machine learning points to an information content in the connectivity patterns that has not yet been captured in conventional network analyses.