Modelling variability in functional brain networks using embeddings

Functional neuroimaging techniques allow us to estimate functional networks that underlie cognition. However, these functional networks are often estimated at the group level and do not allow for the discovery of, nor benefit from, subpopulation structure in the data, i.e. the fact that some recording sessions maybe more similar than others. Here, we propose the use of embedding vectors (c.f. word embedding in Natural Language Processing) to explicitly model individual sessions while inferring dynamic networks across a group. This vector is effectively a "fingerprint" for each session, which can cluster sessions with similar functional networks together in a learnt embedding space. We apply this approach to estimate dynamic functional connectivity, using Hidden Markov Models (HMMs), which are popular methods for inferring dynamic networks, to model individual sessions in neuroimaging data. We call this approach HIVE (HMM with Integrated Variability Estimation). Using simulated data, we show that HIVE can recover the true, underlying inter-session variability and show improved performance over existing approaches. Using real magnetoencephalography data, we show the learnt embedding vectors (session fingerprints) reflect meaningful sources of variation across a population (demographics, scanner types, sites, etc). Overall, HIVE provides a powerful new technique for modelling individual sessions while leveraging information available across an entire group. HighlightsO_LIWe proposed the use of embedding vectors and a novel variability encoding block for inferring individualised brain networks in neuroimaging data. C_LIO_LIWe apply this approach to estimate dynamic functional connectivity using the Hidden Markov Models (HMMs) and explicitly model variability in the training dataset. We call this new model HIVE (HMM with Integrated Variability Estimation) C_LIO_LIWe demonstrate the advantages of HIVE over traditional approaches using both simulated and real MEG data. C_LIO_LIWe show HIVE learns meaningful variability in the data (e.g. measurement site, scanner type, demographics) in an unsupervised manner. C_LIO_LIThe datasets and scripts for performing all the analysis in this paper are made publicly available. C_LI