186 words Main text: 4176 wordsReferences: 36
AbstractThe concept of brain states, functionally relevant large-scale patterns, has become popular in neuroimaging. Not all components of such patterns are equally characteristic for each brain state, but machine learning provides a possibility of extracting the structure of brain states from functional data. However, the characterization in terms of functional connectivity measures varies widely, from cross-correlation to phase coherence, and the idea that different measures will provide the similar information is a common assumption made in neuroimaging. Here, we compare the performance of phase coherence, pairwise covariance, correlation, model-based covariance and model-based precision for a dataset of subjects performing five different cognitive tasks. We employ multinomial logistic regression for classification and consider two types of cross-validation schemes, between-and within-subjects. Furthermore, we investigate whether classification is robust for different temporal window lengths. We find that informative TASK INFORMATION ACROSS FUNCTIONAL CONNECTIVITY METHODS 1 links for the classification, meaning changes between tasks that are consistent across subjects, are entirely uncorrelated between correlation and covariance. These results indicate that the corresponding FC signature can strongly differ across FC methods used and that interpretation is subject to caution in terms of subnetworks related to a task.