Macroscopic quantities of collective brain activity during wakefulness and anesthesia
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
The premise behind this manuscripts is that concepts from thermodynamics and statistical mechanics can be used to understand brain states and the transitions between such states, just like it is done with transitions between solid and liquid states in well define thermodynamic systems. While this is an interesting attempt to use thermodynamic concepts and equations to analyze fMRI signals, the legwork needed to meaningfully translate those concepts to understand their actual physiological meaning and their relationship to brain function has not yet been achieved.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
- Evaluated articles (eLife)
Abstract
The study of states of arousal is key to understand the principles of consciousness. Yet, how different brain states emerge from the collective activity of brain regions remains unknown. Here, we studied the fMRI brain activity of monkeys during wakefulness and anesthesia-induced loss of consciousness. Using maximum entropy models, we derived collective, macroscopic properties that quantify the system’s capabilities to produce work, to contain information and to transmit it, and that indicate a phase transition from critical awake dynamics to supercritical anesthetized states. Moreover, information-theoretic measures identified those parameters that impacted the most the network dynamics. We found that changes in brain state and in state of consciousness primarily depended on changes in network couplings of insular, cingulate, and parietal cortices. Our findings suggest that the brain state transition underlying the loss of consciousness is predominantly driven by the uncoupling of specific brain regions from the rest of the network.
Article activity feed
-
Evaluation Summary:
The premise behind this manuscripts is that concepts from thermodynamics and statistical mechanics can be used to understand brain states and the transitions between such states, just like it is done with transitions between solid and liquid states in well define thermodynamic systems. While this is an interesting attempt to use thermodynamic concepts and equations to analyze fMRI signals, the legwork needed to meaningfully translate those concepts to understand their actual physiological meaning and their relationship to brain function has not yet been achieved.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)
-
Reviewer #1 (Public Review):
This article considers the application of maximum entropy models to the analysis of fMRI data from macaque monkeys under anesthesia using various drugs, and while awake. The authors binarize the raw fMRI data, and use the binarized data to infer networks that underlie maximum entropy models of the binarized data.
The authors argue that the fragmentation of networks, specifically the uncoupling of specific brain regions from others, underlies the transition into loss of consciousness
under anesthesia.The authors use concepts/ideas from statistical mechanics, specifically criticality and supercriticality to to describe the state of the brain during awake to anesthetized states.
Overall comments:
I found the paper well written, organized and motivated from the perspective of the need for network analyses of …
Reviewer #1 (Public Review):
This article considers the application of maximum entropy models to the analysis of fMRI data from macaque monkeys under anesthesia using various drugs, and while awake. The authors binarize the raw fMRI data, and use the binarized data to infer networks that underlie maximum entropy models of the binarized data.
The authors argue that the fragmentation of networks, specifically the uncoupling of specific brain regions from others, underlies the transition into loss of consciousness
under anesthesia.The authors use concepts/ideas from statistical mechanics, specifically criticality and supercriticality to to describe the state of the brain during awake to anesthetized states.
Overall comments:
I found the paper well written, organized and motivated from the perspective of the need for network analyses of brain activity. As I do not have expertise in anesthesia, I will let the other referees comment on the aspects of the paper that deal with its implications on our understanding of anesthesia. My comments will focus on the methodological aspects of this work.
From a methodological perspective, the authors justify the binarization of the fMRI data from existing work, notably references 26 and 27. In my opinion, the strength of the paper lies its ability to show that the binarized representation, after some processing, can serve as a useful feature for classifying different states of the brain. I found this very interesting.
In my opinion, the paper's weakness lies in the fact that I find some of the methodological choices hard to justify. If I grant the authors the binarization of the data, I thought the choice of maximum entropy models, as opposed to other forms of point-process models, such as the ones shared below (other examples abound), requires further justification. Among other things, these models seem to assume independence of binary vectors across times, which I find hard to justify.
-
Reviewer #2 (Public Review):
Ponce-Alvarez et al. investigated the use of concepts from thermodynamics and statistical mechanics to uncover control parameters behind the transitions between brain states. The authors performed fMRI recordings in 5 monkeys using different anesthetics to transition the animals from awake to anesthetized state. Then the authors proceed to binarize the signals using an arbitrary z-scored based threshold, and apply a host of thermodynamic equations to the binarize data in order to identify system control parameters. The authors conclude that the coupling of the binarized activity from each individual ROI to the population activity is related to brain state transitions, just like temperature is related to transition from ice to water.
Major concerns: Statistical mechanics was developed to explain macroscopic …
Reviewer #2 (Public Review):
Ponce-Alvarez et al. investigated the use of concepts from thermodynamics and statistical mechanics to uncover control parameters behind the transitions between brain states. The authors performed fMRI recordings in 5 monkeys using different anesthetics to transition the animals from awake to anesthetized state. Then the authors proceed to binarize the signals using an arbitrary z-scored based threshold, and apply a host of thermodynamic equations to the binarize data in order to identify system control parameters. The authors conclude that the coupling of the binarized activity from each individual ROI to the population activity is related to brain state transitions, just like temperature is related to transition from ice to water.
Major concerns: Statistical mechanics was developed to explain macroscopic behavior of well defined physical systems of atoms or molecules. The concepts were developed to explain how vibrational modes and collisions between atoms/molecules are related to temperature and pressure, and how they can be related to the ability of the system to perform a work. Within that framework, concepts such as entropy, criticality, control parameter, partition function, and Helmholtz free energy (to name some) have a clear physical meaning. I am not sure that the same is true for the brain, unless it was to be explicitly studied in terms of brain temperature. For example, hypothermia produces an EEG very similar to the one observed during deep general anesthesia. While the BOLD signal could be a proxy for brain temperature, the study provides no model to translate such signal to brain temperature. I don't deny that neuron activity is driven by thermodynamic processes coming mostly from the mitochondria. However, a lot of legwork has to be performed before such concepts can be applied to fMRI signals. This painful lesson was learned many years ago in meteorology, a field temperature and pressure measurements are explicit, but where still it is not possible to abstract microscopic changes to macroscopic variables, and computationally intensive finite element modeling must be performed instead. We currently do have models that take into account the explicit structure of the brain, and have the computational power needed to implement it.
Other concerns:
Is binarization necessary? the authors show that similar results could be achieved without binarizing, so the entire paper could have been written without it. Seems to me that binarization of continuous signals is a trick to use only when online or closed loop manipulations are needed.
The actual N of the study is 5, the number of biological units (monkeys). This is not a limitation per se, as many monkey studies only have two animals, but the scans within the same monkey should be considered as repeated measures within each subject, and the authors should use a mixed-effects model to analyze the results, rather than treating each scan as independent. Furthermore, in each graph the data coming from each monkey should be labeled differently. It is also unclear how many monkeys received each of the anesthetic treatments. This should be precisely described, and even if the design is unbalanced, it can be rigorously dealt with by using a mixed-effects model.
The authors report a correlation of 0.3 as being "high". Unlike information-theoretical quantities, statistics like correlation have the advantage of being bounded, in this case between -1 and 1. A correlation value of 0.3 is in itself rather modest, and indicative of a weak correlation in itself, independent of its statistical significance when compared to other measures. Given that the rest of the study relies in this metric, the rest of results are not convincing.
The authors correctly acknowledge that the parameters that control the state transitions are unknown, and state their purpose to find them. However, proper parameter selection it is not performed.
-
