One-shot normative modelling of whole-brain functional connectivity

Many brain diseases and disorders lack objective measures of brain function as indicators of pathology, which has recently spurred the use of normative modelling in neuroimaging. Normative models characterize the normal variation of brain measurements given sex and age, thereby allowing identification of abnormalities as deviations from normal. Normative modelling of brain function is typically based on predicting functional connectivity (FC) between each pair of brain regions. But the human brain is an extremely integrated organ, and brain disease often has widespread effects that are not well captured by piecemeal analyses, i.e. connection by connection. We propose Functional Connectivity Integrative Normative Modelling (FUNCOIN), developing a whole-brain normative model of FC from a large resting-state fMRI data set that captures whole-network-level changes associated with sex and age. This model can significantly, and substantially, uncover abnormal FC patterns in Parkinson’s disease patients even on scans up to 5.5 years before diagnosis.