Intensity / Correlation Thresholding of FMRI Data: Data-driven Regions of Interest using Bridge Voxels

Abstract

Analysis of functional magnetic resonance imaging data considers temporal correlation of hemodynamic intensity patterns to a known activation paradigm as well as spatial regions of similar responses. Default regions of interest are often obtained by analyst-directed thresholding of intensity or correlation values. This paper presents a method to determine data-driven regions of interest using image erosion to identify structurally significant components in the default regions, namely bridge voxels.