Proof-of-concept of a novel structural equation modelling approach for the analysis of functional magnetic resonance imaging data applied to investigate individual differences in human pain responses

A novel network analysis method is demonstrated for applications with functional magnetic resonance imaging (fMRI) data. The method is based on structural equation modeling (SEM) plus modeling of physiological responses in order to explain blood oxygenation-level dependent (BOLD) responses across interconnected regions. The method, termed structural and physiological modeling (SAPM) aims to overcome a weakness of previous analysis methods by estimating both input and output signaling of every region of a network. The results also provide weighting factors (B) which describe the influence of each input signal to a region on its output signaling to another region. The SAPM method is demonstrated by applying it to fMRI data from the brainstem and spinal cord in 55 healthy participants undergoing repeated applications of a heat pain stimulation paradigm. Data are also analyzed using our established SEM method for comparison. The results with both methods indicate that individual differences in nociceptive processing are mediated by differences in descending regulation of spinal cord neurons under the influence of both the nucleus tractus solitarius and periaqueductal gray region. The SAPM results show that BOLD responses in the entire network can be explained during all periods of the stimulation paradigm based on two latent (unobserved) input signaling sources, and a model of the predicted BOLD responses to the heat stimulus. The results demonstrate the concept of our novel SAPM method and provide evidence for its validity. Additional studies are needed to further develop the method and its applications to investigations of complex neural processes across networks.

Automated summary

This paper demonstrates a novel network analysis method for functional magnetic resonance imaging (fMRI) data. The method combines structural equation modeling (SEM) and physiological response modeling to explain interconnected regions' blood oxygenation-level dependent (BOLD) responses. The results show that the method can estimate both input and output signaling of each region and provide weighting factors for the influence of input signals on output signaling. The results also suggest that individual differences in nociceptive processing are mediated by differences in descending regulation of spinal cord neurons under the influence of certain brain regions.