Meeting Banner
Abstract #3364

Assessing Fit of Individuals to Group-Derived Structural Equation Models of Resting-State FMRI Data

George Andrew James1, Richard Cameron Craddock1, Mary Elizabeth Kelley2, Paul E. Holtzheimer3, Boadie Dunlop3, Charles Nemeroff3, Helen S. Mayberg3, Xiaoping Phillip Hu1

1Biomedical Engineering, Emory University / Georgia Institute of Technology, Atlanta, GA, USA; 2Biostatistics and Bioinformatics, Emory University, Atlanta, GA, USA; 3Psychiatry and Behavioral Sciences, Emory University, Atlanta, GA, USA

Structural equation modeling (SEM) of PET data has shown neuroanatomic interactions to predict major depressive disorder (MDD) patients response to therapy. By taking advantage of fMRIs high temporal resolution, we propose to extend this work into the domain of individual patients. Forty-six never treated MDD patients underwent resting-state fMRI scanning. The model best fitting the group was then fit to individual patients. Histogram analysis of individuals path coefficients demonstrate some paths (midanterior cingulate to orbitofrontal) are Gaussian distributed about the group value while others (midanterior cingulate to ventromedial prefrontal) are not, potentially indicating group commonalities and subgroup differences, (respectively).