Predicting Treatment Response from Resting State fMRI Data: Comparison of Parcellation Approaches

Abstract

Resting state fMRI reveals intrinsic network characteristics present in the brain. They are correlated with behavioral measures, and have made surprising insights in the brains' connectivity structure possible. At the core of many of those studies is the correlation of behavioral measures, and the characteristics of networks among a set of brain regions. In this paper we evaluate methods that identify functional networks in resting state fMRI in light of predicting treatment response of patients suffering from social anxiety disorder. Results illustrate differences in prediction when obtaining network labelings by population-wide-clustering, subject-specific parcellation, transferring anatomical region labels, or mapping networks from a previous large scale resting state study.