A Feature-Based Approach To Combine Multimodal Brain Imaging Data
Institute of Living/Yale University
Approaches for combining or fusing data in brain imaging can be put on aspectrum with meta-analysis (highly distilled data) to examine convergentevidence at one end and large-scale computational modeling (highly detailedtheoretical modeling) at the other end. In between are methods that attemptto do direct data fusion. In this study, we present a general ICA fusionframework and introduce a method to assess the value of combining differenttypes of data by using a discrimination metric based on the Kullback-Leibler(KL) divergence to evaluate the joint distributions. The data types we focuson in this paper are fMRI, structural MRI (sMRI), and EEG. We show that bycombining modalities in certain ways, performance (ability to distinguishschizophrenia patients from controls) is improved.