Exploratory structural equation modeling of resting-state fMRI: Applicability of group models to individual subjects

G. Andrew James, Mary E. Kelley, R. Cameron Craddock, Paul E. Holtzheimer, Boadie W. Dunlop, Charles B. Nemeroff, Helen S. Mayberg, Xiaoping P. Hu

Research output: Contribution to journalArticlepeer-review

43 Scopus citations


The extension of group-level connectivity methods to individual subjects remains a hurdle for statistical analyses of neuroimaging data. Previous group analyses of positron emission tomography data in clinically depressed patients, for example, have shown that resting-state connectivity prior to therapy predicts how patients eventually respond to pharmacological and cognitive-behavioral therapy. Such applications would be considerably more informative for clinical decision making if these connectivity methods could be extended into the individual subject domain. To test such an extension, 46 treatment-naïve depressed patients were enrolled in an fMRI study to model baseline resting-state functional connectivity. Resting-state fMRI scans were acquired and submitted to exploratory structural equation modeling (SEM) to derive the optimal group connectivity model. Jackknife and split sample tests confirm that group model was highly reproducible, and path weights were consistent across the best five group models. When this model was applied to data from individual subjects, 85% of patients fit the group model. Histogram analysis of individual subjects' paths indicate that some paths are better representative of group membership. These results suggest that exploratory SEM is a viable technique for neuroimaging connectivity analyses of individual subjects' resting-state fMRI data.

Original languageEnglish (US)
Pages (from-to)778-787
Number of pages10
Issue number3
StatePublished - Apr 15 2009
Externally publishedYes

ASJC Scopus subject areas

  • Neurology
  • Cognitive Neuroscience


Dive into the research topics of 'Exploratory structural equation modeling of resting-state fMRI: Applicability of group models to individual subjects'. Together they form a unique fingerprint.

Cite this