Author Summary The discovery of resting-state networks has greatly influenced the investigation of brain functioning, shifting the focus from local regions involved in cognitive tasks to the ongoing spontaneous dynamics in global networks. This research goes beyond that shift and proposes investigating how human cognition is shaped by the interactions between whole-brain networks embedded in a low-dimensional manifold space. To achieve this, a combination of deep variational autoencoders with computational modelling is used to construct a dynamic model of brain networks, fitted to whole-brain dynamics measured with functional magnetic resonance imaging (fMRI). The results show that during cognitive tasks, highly flexible reconfigurations of task-driven network interaction patterns occur, and these patterns, in turn, can be used to accurately classify different cognitive tasks. Importantly, using this low-dimensional whole-brain network model provides significantly better results than working in the conventional brain space.

Different whole-brain computational models have been recently developed to investigate hypotheses related to brain mechanisms. Among these, the Dynamic Mean Field (DMF) model is particularly attractive, combining a biophysically realistic model that is scaled up via a mean-field approach and multimodal imaging data. However, an important barrier to the widespread usage of the DMF model is that current implementations are computationally expensive, supporting only simulations on brain parcellations that consider less than 100 brain regions. Here, we introduce an efficient and accessible implementation of the DMF model: the FastDMF. By leveraging analytical and numerical advances—including a novel estimation of the feedback inhibition control parameter and a Bayesian optimization algorithm—the FastDMF circumvents various computational bottlenecks of previous implementations, improving interpretability, performance, and memory use. Furthermore, these advances allow the FastDMF to increase the number of simulated regions by one order of magnitude, as confirmed by the good fit to fMRI data parcellated at 90 and 1,000 regions. These advances open the way to the widespread use of biophysically grounded whole-brain models for investigating the interplay between anatomy, function, and brain dynamics and to identify mechanistic explanations of recent results obtained from fine-grained neuroimaging recordings.

Author Summary Significant advances in nonlinear dynamics and computational modeling have opened up the possibility of studying how whole-brain dynamics may be impacted by traumatic brain injury (TBI). One hypothesis is that whole-brain dynamics may show differential spatiotemporal patterns during the recovery trajectory. We used a novel turbulence framework to examine how several measures related to turbulent dynamics differ between healthy controls and TBI patients at 3, 6, and 12 months post-injury. This framework revealed a significant reduction in these empirical measures after TBI differentially affecting long distances, with the largest change at 6 months post-injury. The alterations observed at network level, however, showed a certain degree of recovery after 1 year. In addition, the Hopf whole-brain model demonstrated decreased susceptibility and information-encoding capability after TBI. The clinical implications of this work are discussed. Abstract It has been previously shown that traumatic brain injury (TBI) is associated with reductions in metastability in large-scale networks in resting-state fMRI (rsfMRI). However, little is known about how TBI affects the local level of synchronization and how this evolves during the recovery trajectory. Here, we applied a novel turbulent dynamics framework to investigate whole-brain dynamics using an rsfMRI dataset from a cohort of moderate to severe TBI patients and healthy controls (HCs). We first examined how several measures related to turbulent dynamics differ between HCs and TBI patients at 3, 6, and 12 months post-injury. We found a significant reduction in these empirical measures after TBI, with the largest change at 6 months post-injury. Next, we built a Hopf whole-brain model with coupled oscillators and conducted in silico perturbations to investigate the mechanistic principles underlying the reduced turbulent dynamics found in the empirical data. A simulated attack was used to account for the effect of focal lesions. This revealed a shift to lower coupling parameters in the TBI dataset and, critically, decreased susceptibility and information-encoding capability. These findings confirm the potential of the turbulent framework to characterize longitudinal changes in whole-brain dynamics and in the reactivity to external perturbations after TBI.

Author Summary Here we used a thermodynamics-inspired, deep learning based Temporal Evolution NETwork (TENET) framework to assess the asymmetry in the flow of events, ‘arrow of time’, in human brain signals. This was applied to large-scale HCP neuroimaging data which showed significant changes between the hierarchy of orchestration for the resting state and seven different cognitive tasks. Similarly, the hierarchy of the resting state is significantly different in health compared to neuropsychiatric disorders. This framework provided new insights into the orchestrating of brain dynamics in different brain states. Abstract A promising idea in human cognitive neuroscience is that the default mode network (DMN) is responsible for coordinating the recruitment and scheduling of networks for computing and solving task-specific cognitive problems. This is supported by evidence showing that the physical and functional distance of DMN regions is maximally removed from sensorimotor regions containing environment-driven neural activity directly linked to perception and action, which would allow the DMN to orchestrate complex cognition from the top of the hierarchy. However, discovering the functional hierarchy of brain dynamics requires finding the best way to measure interactions between brain regions. In contrast to previous methods measuring the hierarchical flow of information using, for example, transfer entropy, here we used a thermodynamics-inspired, deep learning based Temporal Evolution NETwork (TENET) framework to assess the asymmetry in the flow of events, ‘arrow of time’, in human brain signals. This provides an alternative way of quantifying hierarchy, given that the arrow of time measures the directionality of information flow that leads to a breaking of the balance of the underlying hierarchy. In turn, the arrow of time is a measure of nonreversibility and thus nonequilibrium in brain dynamics. When applied to large-scale Human Connectome Project (HCP) neuroimaging data from close to a thousand participants, the TENET framework suggests that the DMN plays a significant role in orchestrating the hierarchy, that is, levels of nonreversibility, which changes between the resting state and when performing seven different cognitive tasks. Furthermore, this quantification of the hierarchy of the resting state is significantly different in health compared to neuropsychiatric disorders. Overall, the present thermodynamics-based machine-learning framework provides vital new insights into the fundamental tenets of brain dynamics for orchestrating the interactions between cognition and brain in complex environments.

Large variability exists across brain regions in health and disease, considering their cellular and molecular composition, connectivity, and function. Large-scale whole-brain models comprising coupled brain regions provide insights into the underlying dynamics that shape complex patterns of spontaneous brain activity. In particular, biophysically grounded mean-field whole-brain models in the asynchronous regime were used to demonstrate the dynamical consequences of including regional variability. Nevertheless, the role of heterogeneities when brain dynamics are supported by synchronous oscillating state, which is a ubiquitous phenomenon in brain, remains poorly understood. Here, we implemented two models capable of presenting oscillatory behavior with different levels of abstraction: a phenomenological Stuart–Landau model and an exact mean-field model. The fit of these models informed by structural- to functional-weighted MRI signal (T1w/T2w) allowed us to explore the implication of the inclusion of heterogeneities for modeling resting-state fMRI recordings from healthy participants. We found that disease-specific regional functional heterogeneity imposed dynamical consequences within the oscillatory regime in fMRI recordings from neurodegeneration with specific impacts on brain atrophy/structure (Alzheimer’s patients). Overall, we found that models with oscillations perform better when structural and functional regional heterogeneities are considered, showing that phenomenological and biophysical models behave similarly at the brink of the Hopf bifurcation. Author Summary Significant progress has been made in understanding the effects of regional heterogeneity on whole-brain dynamics. With imaging technologies, the number of high-resolution reference maps of brain structure and function has been increased, and whole-brain computational models have provided a suitable avenue to investigate the mechanisms supporting the relations between these maps and whole-brain dynamics. Here, we investigate the role of the heterogeneities when synchronous behavior is present in brain dynamics, which we could represent by models capable of oscillating in the presence of a Hopf bifurcation. We found that models with oscillations more faithfully reproduce empirical properties when structural and functional regional heterogeneities are considered, showing that both phenomenological and biophysical models behave similarly at the brink of the Hopf bifurcation.

Psychedelic drugs show promise as safe and effective treatments for neuropsychiatric disorders, yet their mechanisms of action are not fully understood. A fundamental hypothesis is that psychedelics work by dose-dependently changing the functional hierarchy of brain dynamics, but it is unclear whether different psychedelics act similarly. Here, we investigated the changes in the brain’s functional hierarchy associated with two different psychedelics (LSD and psilocybin). Using a novel turbulence framework, we were able to determine the vorticity, that is, the local level of synchronization, that allowed us to extend the standard global time-based measure of metastability to become a local-based measure of both space and time. This framework produced detailed signatures of turbulence-based hierarchical change for each psychedelic drug, revealing consistent and discriminate effects on a higher level network, that is, the default mode network. Overall, our findings directly support a prior hypothesis that psychedelics modulate (i.e., “compress”) the functional hierarchy and provide a quantification of these changes for two different psychedelics. Implications for therapeutic applications of psychedelics are discussed. Author Summary Significant progress has been made in understanding the effects of psychedelics on brain function. One of the main hypotheses is that psychedelics work by changing the functional hierarchy of brain dynamics in a dose-dependent manner, modulating the encoding of the precision of priors, beliefs, or assumptions in the brain. We used a novel turbulence framework to investigate the changes in the brain’s functional hierarchy associated with two different psychedelics (LSD and psilocybin). This framework produced detailed signatures of turbulence-based hierarchical change for each psychedelic drug, revealing consistent and discriminate effects on a higher level network, that is, the default mode network.