Validation of Structural Equation Modeling Methods for Functional MRI Data Acquired in the Human Brainstem and Spinal Cord.

Structural equation modeling (SEM) provides a means of investigating relationships between blood oxygenation level-dependent (BOLD) signal changes in functional MRI data across neuroanatomical regions. The objectives of this study were to demonstrate adapted SEM methods for the brainstem and spinal cord, validate statistical methods and appropriate statistical thresholds, and test the methods with existing data. SEM methods were applied using an anatomical model of regions of the thalamus, brainstem, and spinal cord that are involved with pain processing. Statistical distributions (Z-scores), significance thresholds, and corrections for multiple comparisons were determined from repeated simulations using "null" data sets. SEM analyses were then applied to data from prior studies involving noxious stimulation in healthy participants. Z-score distributions were observed to vary with the number of source regions modeled, the number of time points (volumes) included in the analysis, and the time span (epoch) used for dynamic analyses. Appropriate choices of statistical thresholds and corrections for multiple comparisons were demonstrated. The results reveal consistent network features across/within studies, as well as dependences on study conditions. They show the effectiveness of a SEM method for functional MRI data from the brainstem and spinal cord.