Mixtures of large-scale dynamic functional brain network modes

Gohil, Chetan; Roberts, Evan; Timms, Ryan C; Skates, Alex; Higgins, Cameron; Quinn, Andrew; Pervaiz, Usama; Amersfoort, Joost van; Notin, Pascal; Gal, Yarin; Adaszewski, Stanislaw; Woolrich, Mark W.

doi:10.1016/j.neuroimage.2022.119595

Cited by 18 publications

(49 citation statements)

References 84 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DyNeMo does have the additional advantage of using a richer temporal regularisation through the use of a deep recurrent network. This has been shown to capture longer range temporal structure than the HMM [26], and exploring the cognitive importance of long-range temporal structure is an interesting area of future investigation [52]. It is possible to quantify which model is better using a downstream task.…”

Section: Discussionmentioning

confidence: 99%

“…These are models that learn the probability distribution of the training data. In this report, we will focus on two generative models for time series data: the Hidden Markov Model (HMM) [34] and Dynamic Network Modes (DyNeMo) [26]. Both of these models (discussed further below) incorporate an underlying dynamic latent variable in the generative process.…”

Section: Generative Modelsmentioning

confidence: 99%

“…Alternatively, if we were interested in calculating the same summary statistics as the HMM (fractional occupancy, lifetime, interval, switching rate) we would first need to binarise the mode time courses. This can be done using a two-component Gaussian Mixture Model (GMM), which is discussed in [26]. Note, an additional complication related to the mode time course is that it does not contain any information regarding the relative magnitude of each mode covariance.…”

Section: Ae-hmmmentioning

confidence: 99%

“…Here, we present the OHBA Software Library Dynamics Toolbox (osl-dynamics), a Python package that meets two far-reaching methodological challenges that limit the field of cognitive neuroscience: burst detection and the identification of dynamic functional brain networks. It does so by deploying data driven generative models that have a proven ability to adapt to the data from a wide range of imaging modalities, and can learn the spatio-temporal characteristics of brain activity, with few assumptions and at fast timescales [24,25,26].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

osl-dynamics: A toolbox for modelling fast dynamic brain activity

Gohil

Huang

Roberts

et al. 2023

Preprint

View full text Add to dashboard Cite

Neural activity contains rich spatio-temporal structure that corresponds to cognition. This includes oscillatory bursting and dynamic activity that span across networks of brain regions, all of which can occur on timescales of a tens of milliseconds. While these processes can be accessed through brain recordings and imaging, modelling them presents methodological challenges due to their fast and transient nature. Furthermore, the exact timing and duration of interesting cognitive events is often a priori unknown. Here we present the OHBA Software Library Dynamics Toolbox (osl-dynamics), a Python-based package that can identify and describe recurrent dynamics in functional neuroimaging data on timescales as fast as tens of milliseconds. At its core are machine learning generative models that are able to adapt to the data and learn the timing, as well as the spatial and spectral characteristics, of brain activity with few assumptions.osl-dynamicsincorporates state-of-the-art approaches that can be, and have been, used to elucidate brain dynamics in a wide range of data types, including magneto/electroencephalography, functional magnetic resonance imaging, invasive local field potential recordings and electrocorticography. It also provides novel summary measures of brain dynamics that can be used to inform our understanding of cognition, behaviour and disease. We hopeosl-dynamicswill further our understanding of brain function, through its ability to enhance the modelling of fast dynamic processes.HighlightsAn open-source toolbox for identify and describing brain dynamics in neuroimaging data on fast timescales.Includes visualisation and quantification of oscillatory bursting and network dynamics.Provides novel summary measures and group analysis tools of brain dynamics that can be used to inform our understanding of cognition, behaviour and disease.Implemented in Python and makes use ofTensorFlow.Includes comprehensive documentation and tutorials.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Generative Modelsmentioning

confidence: 99%

Section: Ae-hmmmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

osl-dynamics: A toolbox for modelling fast dynamic brain activity

Gohil

Huang

Roberts

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…DyNeMo. Here, the model assumes the data is generated by a set of modes [11], each with its own characteristic covariance matrix. Crucially, these modes can overlap.…”

Section: Dynamic Network Modelsmentioning

confidence: 99%

Dynamic Network Analysis of Electrophysiological Task Data

Gohil,

Kohl,

Huang

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

An important approach for studying the human brain is to use functional neuroimaging combined with a task. In electrophysiological data this often involves a time-frequency analysis, in which recorded brain activity is time-frequency transformed and epoched around task events of interest, followed by trial-averaging of the power. Whilst this simple approach can reveal fast oscillatory dynamics, the brain regions are analysed one at a time. This causes difficulties for interpretation and a debilitating number of multiple comparisons. In addition, it is now recognised that the brain responds to tasks through the coordinated activity of networks of brain areas. As such, techniques that take a whole-brain network perspective are needed. Here, we show how the oscillatory task responses from conventional time-frequency approaches, can be represented more parsimoniously at the network level using two state-of-the-art methods: the HMM (Hidden Markov Model) and DyNeMo (Dynamic Network Modes). Both methods reveal frequency-resolved networks of oscillatory activity with millisecond resolution. Comparing DyNeMo, HMM and traditional oscillatory response analysis, we show DyNeMo can identify task activations/deactivations that the other approaches fail to detect. DyNeMo offers a powerful new method for analysing task data from the perspective of dynamic brain networks.

show abstract

Comparison between EEG and MEG of static and dynamic resting‐state networks

Cho,

van Es,

Woolrich

et al. 2024

Human Brain Mapping

View full text Add to dashboard Cite

The characterisation of resting‐state networks (RSNs) using neuroimaging techniques has significantly contributed to our understanding of the organisation of brain activity. Prior work has demonstrated the electrophysiological basis of RSNs and their dynamic nature, revealing transient activations of brain networks with millisecond timescales. While previous research has confirmed the comparability of RSNs identified by electroencephalography (EEG) to those identified by magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI), most studies have utilised static analysis techniques, ignoring the dynamic nature of brain activity. Often, these studies use high‐density EEG systems, which limit their applicability in clinical settings. Addressing these gaps, our research studies RSNs using medium‐density EEG systems (61 sensors), comparing both static and dynamic brain network features to those obtained from a high‐density MEG system (306 sensors). We assess the qualitative and quantitative comparability of EEG‐derived RSNs to those from MEG, including their ability to capture age‐related effects, and explore the reproducibility of dynamic RSNs within and across the modalities. Our findings suggest that both MEG and EEG offer comparable static and dynamic network descriptions, albeit with MEG offering some increased sensitivity and reproducibility. Such RSNs and their comparability across the two modalities remained consistent qualitatively but not quantitatively when the data were reconstructed without subject‐specific structural MRI images.

show abstract

Mixtures of large-scale dynamic functional brain network modes

Cited by 18 publications

References 84 publications

osl-dynamics: A toolbox for modelling fast dynamic brain activity

osl-dynamics: A toolbox for modelling fast dynamic brain activity

Dynamic Network Analysis of Electrophysiological Task Data

Comparison between EEG and MEG of static and dynamic resting‐state networks

Contact Info

Product

Resources

About