Merging of structure & function in The Virtual Brain. Top: structural neuroimaging, below: functional neuroimaging
Unraveling the keys to the brain
What makes The Virtual Brain unique is a rather new way of addressing the inherent difficulties of simulating a large-scale network like the human brain:
-
Understanding the brain's behavior as a network's performance at all seems obvious but isn't so much in hindsight.
The Virtual Brain builds upon the discovery of the critical network parameters of the human brain, their influence to functional processes and their proper tweaking to rectify a malfunctioning or damaged network.
- Rather than making simplifying assumptions about topology, density and range of large-scale connectivity (anatomical realism), The Virtual Brain will invoke the Connectome, simultaneously integrating multiple modes of network activity.
- The Virtual Brain purposes to include an array of new and useful measures for the brain's organization thanks to extensive use of graph theory: segregation, integration, efficiency and influence of subnetworks, nodes and their edges.
- For the first time, The Virtual Brain will provide the same qualities and quantifications of common neuro-imaging methods (EEG, MEG, fMRI) as a real brain, making it ideal for experimental validation and customization.
Realizing that network nodes in actual human beings are far from homogenous, The Virtual Brain captures the functioning of the sub-networks of the human brain through the novel concept of the space-time structure of the network couplings featuring means for quantifiable coupling matrices within and across regions.
Upcoming versions will constantly process these experimental results through machine learning methods, further refining The Virtual Brain to the best match for an individual clinical case.
While the responses of isolated brain regions are well studied today, The Virtual Brain overlays an elaborate map of functional pathways on and between these regions:
A robust mathematical core of anatomically realistic connection matrices (based on DTI/DSI scans) defines the network itself, a physiological model of neural populations captures the individual regions. This model is used to search for the crucial points in actual network damage scenarios in brain health.
Biologically realistic cortical connectivity of brain areas
Accounting for differing speeds
While the brain's well-known folded structure was always admired as nature's masterpiece of squeezing as many neurons as possible in a constrained space, the second anatomical feat of this structure was discovered only recently.
Turns out that the distance a neural signal has to travel is absolutely necessary to truly understand the brain‘s network: only when taking into account the unique time-delays for neuron-to-neuron signal transmission stemming from the brain's anatomy, one can successfully emulate the brain's measurable spatial-temporal patterns.
Without those innocuous time-delays (ranging from few to hundreds of milliseconds, depending on distance and direction), our brain would simply cease to work correctly.
Resting is the new active
Historically, the human brain was understood to behave like a classic feed-forward information processing system: you exhibit it to an external stimulus (sensory input) or let it steer a task (motoric action) and its measurable signaling patterns will reflect this behaviour – somehow.
What's missing from this picture is a reasonable explanation of the undoubtedly ongoing brain activity when it's doing just ... nothing. Is it only noise, irrelevant for functional processes or rather something else?
The fascinating results from a large number of experimental investigations in the last decade point to "something else":
Recent theoretical analysis has indeed confirmed the major role of the so-called "resting state networks" within the brain. Their complex oscillations on different time-scales even provide the utter foundation of functional processes within the brain.
Metaphorically speaking, the resting state of the brain can be pictured as a nimble, always vigilant tennis player, waiting on his baseline for the new service of his opponent (which would be an outside stimulus or a task being performed). Thanks to his constant motion, mindfully envisioning possible routes, he can react more readily to events from various directions.
The autonomy of scales
Traditional understanding of the brain divides it into different regions and lobes of varying size (from tiny to huge), each being responsible for some different cognitive abilities. Additionally, we know that the brain's signaling patterns show oscillations on different time-scales such as 10 Hz is involved in the resting brain, 4 Hz in memory and 40 Hz in the cognitive brain.
Unifying the observed behaviour and organization of the brain on multiple spatio-temporal scales, several approaches have been studied by perusing analogies from known physical phenomena (e.g. fluids or magnetization). However, none of them could explain, let alone model the brain's behaviour in a sufficiently solid way.
The Virtual Brain is well underway to invoke a clever blend between classic unifying multiscale frameworks and pyramid-style approaches seen in other informatics-oriented projects:
- Moving upwards through the scales, The Virtual Brain uses dimension reduction techniques to keep the principal influence of smaller scales on larger ones while leveling out their inherent complexity.
- Moving downwards through the scales, more detailed modeling parameters can be used, e.g. to test specific hypotheses.
- No particular scale is dominating the model. Instead, multiple scales operate through mutual interdependence, which has also beneficial effects on the computational load of the model.
Multiscale simulation results from the resting state (see Deco, Jirsa, McIntosh 2011)