Virtual deep brain stimulation: Multiscale co-simulation of a spiking basal ganglia model and a whole-brain mean-field model with The Virtual Brain

The quest for multiscale brain modeling

“…Third, the performance of TVB can be enhanced by increasing the number of nodes toward a fine-grained representation of brain properties. Finally, more precise representations of local neural activities can be obtained substituting NM with MF models [108] or, in perspective, with SNNs [109][110][111]. This means, de facto, that a bottom-up strategy can be used to obtain the model to be inserted into the top-down flow, thereby generating a hybrid simulation on mixed scales (cf Figures 2 and 3).…”

“…Unavoidably, TVB at mixed scales poses a greater challenge for humans than for mice [112] because data about mouse neurons and connectivity are far more advanced [14,[18][19][20][21] and allow, in principle, reconstruction of a fine-grained brain model (Figure 3) by leveraging biophysically detailed neuron and microcircuit models [66], SNNs [78], axonal tractography [18], and a large set of multiscale data for reconstruction and validation [3,49]. In principle, a hybrid TVB [109][110][111] would allow an understanding of single neuron dynamics in the whole-brain context and exploration of the impact of single neurons and single spikes on brain dynamics [40][41][42]44,45].…”

“…In essence, virtual brain modeling allows the extraction of hidden information contained in the correlation between structural and functional connectivity through the simulation of brain dynamics [1]. An interesting attempt to bridge scales by inserting microscale into macroscale models has been reported to simulate the effects of deep brain stimulation in which a spiking basal ganglia model has been cosimulated with a whole brain mean-field model using TVB [109][110][111]. Hybrid closed-loop controllers are another category of models that allow cosimulations by embedding bottom-up microscale models into a top-down macroscale architecture.…”

“…In this sense, hybrid robotic controllers are emerging as a sophisticated tool for investigating the neural correlates of behavior (Box 2). The process might be brought forward by combining SNNs [109][110][111] and functional modules [122] with virtual brains. Further, some molecular properties (e.g., extracted from genomic, proteomic, or metabolomic analyses) can be placed in the virtual space of brain atlases and remapped onto virtual brains and local microcircuits, that allowing model fitness to be improved by importing low-level biological features into high-level constructs [106].…”

“…Modeling the relationship between microscopic phenomena and large-scale brain functions could allow us to predict how a drug that binds to specific receptors modifies local and distributed circuit activity, or how genetic alterations to membrane ion channels or receptors will modulate brain functions and dynamics. This in turn would allow the identification of potential targets for pharmacological and physical therapyfor example through electrical or magnetic stimulation of specific circuitsor precision surgery [108][109][110][111][134][135][136] precision medicine (see Outstanding questions), for example the generation of brain digital twins [22,133]. These aim to be personalized replicas of a subject's brain that can be used to simulate specific functionalities and anticipate the consequences of, for example, neurorehabilitation or surgical intervention.…”