Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.

Fast Brain Network Dynamics

Fast transient networks
Fast transient networks in spontaneous human brain activity captured using a Hidden Markov Model

Hidden Markov Models

We are developing novel techniques using methods from machine learning, such as Hidden Markov Models (HMMs), that make it possible to identify dynamic brain networks and dynamic functional connectivity in neuroimaging data at a wide range of time scales. These techniques have also been adapted to work on big population data [1].

Applied to MEG data, this has led to the identification of resting state networks in MEG at 100 ms time scales, orders of magnitude faster than has been shown previously [2]. Using fMRI data, we have shown that transitions between different brain networks are not random, and instead are hierarchically organised in a manner that predicts behaviour [3].

We have also shown that the occurrence of brain network states in MEG coincides with synchronised oscillatory activity across distributed brain networks, consistent with the idea that they provide a mechanism for organising communication across the brain [4].

The occurrence of brain network states has also been used to predict different psychiatric and neurological conditions, e.g. multiple sclerosis [5] and Alzheimer’s Disease [6]; to predict task conditions [7]; and to predict occurrence of spontaneous replay of recently acquired information [8].

Tools for performing these kinds of HMM analyses are available to download and use at the OHBA Analysis Group Software Page (using the HMM-MAR toolbox).

Deep Learning Models

We also develop models using tools from deep learning, such as recurrent neural networks (RNNs). These models can overcome some of the limitations of HMMs, such as the short memory (the Markovian constraint) and mutual exclusivity of states.

A model we are developing is the Variational RNN Auto-Decoder (VRAD). This model is inspired by a popular deep learning framework known as the variational autoencoder. VRAD uses amortised inference to learn a hidden state description of neuroimaging data. Each state represents a large-scale brain network. The temporal dynamics of state switches are captured by an RNN.

Another model we are developing is the Multi-dynamic Adversarial Generator-Encoder (MAGE). This model uses generative adversarial networks to study functional neuroimaging data. MAGE also learns a hidden state description of the data. However, it can also capture different types of state dynamics simultaneously [9].

One puzzling aspect of research into time-varying functional connectivity (FC) has been that it appears to be so stable over time when using techniques like sliding window correlations or (to a lesser extent) the HMM. We show using MAGE that this apparent stability is caused by dynamics in the FC being confounded by dynamics in the mean activity levels. MAGE’s multi-dynamic ability allows changes in the FC and in the mean activity to occur at different times to each other, revealing much stronger changes in FC over time [9].

 

References

    1. Vidaurre D, Abeysuriya R, Becker R, Quinn A … Woolrich M. Discovering dynamic brain networks from big data in rest and task. NeuroImage, 2018.
    2. Baker AP, Brookes MJ, Rezek IA, Smith SM, Behrens T, Probert Smith PJ, Woolrich M. Fast transient networks in spontaneous human brain activity. Elife 2014.
    3. Vidaurre, D., Smith, S. M., Woolrich, M. Brain Network Dynamics are Hierarchically Organised in Time. PNAS 2017. 
    4. Vidaurre D, Hunt L, Quinn AJ, Hunt BAE, Brookes MJ, Nobre AC, Woolrich M.  Spontaneous cortical activity transiently organises into frequency specific phase-coupling networks. Nature Communications. 2018.
    5. Van Schependom J, Vidaurre D, Costers L, Sjogard M, Sima D, Smeets D, D'Hooghe M, D'Haeseleer M, Deco G, Wens V, De Tiege X, Goldman S, Woolrich M, Nagels G. Reduced brain integrity slows down and increases low alpha power in multiple sclerosis. Multiple Sclerosis Journal, 2020.
    6. Sitnikova T, Hughes J, Ahlfors S, Woolrich M, Salat D. Short timescale abnormalities in the states of spontaneous synchrony in the functional neural networks in Alzheimer's disease. NeuroImage: Clinical, 2018.
    7. Quinn A, Vidaurre D, Abeysuriya R, Becker R… Woolrich M. Task-evoked dynamic network analysis through hidden markov modelling. Frontiers in Neuroscience, 2018.
    8. Higgins C, Liu Y, Vidaurre D, Kurth-Nelson Z, Dolan R, Behrens T, Woolrich M. Replay bursts in humans coincide with activation of the default mode and parietal alpha networks. Neuron, 2021.
    9. Pervaiz U, Vidaurre D, Gohil C, Smith S, Woolrich M. Multi-dynamic Modelling Reveals Strongly Time-varying Resting fMRI Correlations. In submission.