Warnings and caveats in brain controllability

Abstract

A recent article by Gu et al. (Nat. Commun. 6, 2015) proposed to characterize brain networks, quantified using anatomical diffusion imaging, in terms of their "controllability", drawing on concepts and methods of control theory. They reported that brain activity is controllable from a single node, and that the topology of brain networks provides an explanation for the types of control roles that different regions play in the brain. In this work, we first briefly review the framework of control theory applied to complex networks. We then show contrasting results on brain controllability through the analysis of five different datasets and numerical simulations. We find that brain networks are not controllable (in a statistical significant way) by one single region. Additionally, we show that random null models, with no biological resemblance to brain network architecture, produce the same type of relationship observed by Gu et al. between the average/modal controllability and weighted degree. Finally, we find that resting state networks defined with fMRI cannot be attributed specific control roles. In summary, our study highlights some warning and caveats in the brain controllability framework.

Neurofeedback: Principles, appraisal, and outstanding issues

“…This issue is especially relevant for DBS devices, but it also indirectly applies to NF, and may have important implications for the determination of neural targets and, more specifically, for their anatomical locality. Remarkably, contrary to previous reports suggesting that brain resting activity may be controllable through a single node representing a given brain region (Gu et al., ), a recent study (Tu et al., ) suggested that even though brain networks might theoretically be structurally controllable, in practice the energy needed to control the system may be disproportionately high to achieve control.…”

“…Indeed, the theoretical network control framework has started being used in neuroscience (see Bassett et al., for a review). However, these studies make rather unrealistic hypotheses on brain activity (Tu et al., ): brain resting dynamics is described in terms of a set of differential equations linearized around a dominant fixed point, an assumption that may account for only limited portions of the phase space, and connectivity dynamics is assumed to be linear time‐invariant (Kim et al., ). Furthermore, numerical control has been shown to fail in practice even for linear systems (Sun & Motter, ): control trajectories turn out to be non‐local in the phase space, that is, the target state may lie very close to the initial condition in the phase space but moving the system from the former to the latter may still require a very long path (see Figure ); there is also a non‐locality trade‐off whereby either the control trajectory is non‐local in the phase space or the control inputs are non‐local in the network (Sun & Motter, ).…”

Driving the brain towards creativity and intelligence: A network control theory analysis

“…Finally, the ability of specific nodes in a network (such as anatomical brain networks) to drive the system into a specific state has been recently questioned (Menara, Gu, Bassett, & Pasqualetti, 2017; Tu et al, 2017). In this paper, we were interested in investigating the theoretical notion of network control theory and individual differences in creativity, without committing to linking between cognitive control processes and network controllability.…”

Structural controllability predicts functional patterns and brain stimulation benefits associated with working memory

“…While applications of control theory in the brain have gained substantial interest in the past few years, the validity of this framework has generated a rigorous theoretical debate Tu et al, 2018), and questions remain about both the foundational tenets of the approach (e.g., Are brain regions controllable?) and the specific manner in which control is executed in the brain.…”

“…As noted above, MC has been inferred to quantify the capacity of a single node to drive the system to difficult-to-reach states. While this inference is driven largely by the frontoparietal distribution of MC (Gu et al, 2015), there is currently little concrete data supporting this interpretation (Tu et al, 2018). Nonetheless, several studies have examined the potential for MC to predict functional patterns through in silico experiments (Betzel et al, 2016;Muldoon et al, 2016) but the identification of a reliable, significant association between MC and functional activity is still lacking (Cornblath et al, 2019), and a systematic investigation into the interdependencies between structural network controllability, functional activity, and cognition remains critical to address the value of this novel network control measure.…”