An Evaluation of Sparse Inverse Covariance Models for Group Functional Connectivity in Molecular Imaging

Evaluating the functional relationships between brain regions measured with neuroimaging provides insight into how the brain is sharing information at a macro scale. Many functional connectivity methods have been developed for dynamic imaging modalities such as functional MRI (fMRI), while less work has focused on models for static molecular imaging techniques such as FDG-PET and Tc-99m HMPAO SPECT across groups of individuals. In this work we provide a quantitative assessment of how well three functional connec- tivity models based on sparse inverse covariance estimation can accurately recover gold standard connectivity patterns across multiple cohorts and data set sizes. We compare the accuracies of learning regularized inverse covariance matrices across cohorts independently with those learned using two different group-based regular- ization models. By using large cohorts of SPECT scans, we are able to provide a quantitative assessment of the accuracy of the models in recovering the gold standard functional conn