What can the brain’s connectome harmonics tell us about consciousness?
A conversation with Dr. Selen Atasoy
Selen Atasoy is a small and intense woman of Turkish descent. She speaks perfect English with a delightful Turkish accent. She is a formidable presenter and speaker and even though she is still young it seems all but assured that she will be a great influencer and scholar in her field of neuroscience for decades to come.
We met at the 2013 Association for the Scientific Study of Consciousness conference in San Diego. We had a fun-filled evening at a restaurant with the developer of the well-known Integrated Information Theory, Giulio Tononi, Giulio’s partner, Chiara Cirelli, and a couple of other people, that involved shooting oysters, eating great food, and not talking much about neuroscience at all!
But there were many other conversations about neuroscience and the mind after that, with both Tononi and Atasoy, as we continued to meet at various conferences and maintain some correspondence. Atasoy’s big breakthrough came with a 2016 paper in Nature Communications, one of the top science journals, on her work on the connectome harmonics that shows key patterns (harmonics) occurring throughout nature (like the fossil trilobite above) also underlie brain function.
She has since followed up this groundbreaking work with many more papers and she continues to flesh out the details of how our brains and minds work, studying the neural signatures of various states of consciousness ranging from psychedelic states, meditation, anesthesia and various disorders of consciousness. She is currently a postdoctoral fellow at the Hedonia Transnational Research Group at Oxford University in England. She pursued her Ph.D jointly at Imperial College London and the Technical University of Munich.
We conducted the following interview by email during 2020.
What is the human “connectome”?
The human ‘connectome’ is a map — a comprehensive structural description — of all neural connections in the human brain. So it is a network description of the human brain’s structural connectivity that not only accounts for individual elements but also their interconnections.
Depending on the scale of description, the individual elements of the connectome can be chosen to represent different functional areas (macro-scale), neural populations (meso-scale), or individual neurons (micro-scale). The connections between individual elements can also be studied at the microscopic scale such as the set of synaptic connections between individual neurons, or at the macroscopic scale such as the white-matter fiber tracts (pathways) between different functional areas.
Given the human brain’s immense complexity, today we do not yet have the technology to create a full-scale human connectome at the micro-scale that describes all of the connections between each pair of individual neurons. However, using various neuroimaging technologies, we are currently able to track the white-matter connections between different parts of the human brain.
In the connectome-harmonic framework [that we have developed], we try to bridge the gap between the macro-scale and meso-scale, as much as possible, by combining the gray matter and white matter connectivity without using any parcellation of the human brain into its functional areas.
This connectome-harmonic framework gives us the opportunity to describe the structural connectivity of the human brain at the highest resolution available using non-invasive neuroimaging technologies such as magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI).
So the connectome-harmonic framework is mapping the brain’s electromagnetic wave patterns and connections rather than the neuroanatomy of synaptic and white-matter connections that the connectome projects have focused on to date?
Yes, the connectome harmonics research that we do is a kind of prediction of the building blocks of the brain’s activity or electromagnetic wave patterns. Just like the different combinations of chemical bases in DNA comprise the fundamental elements of different genes or elementary particles form the building blocks of various atoms and molecules, connectome harmonics are what we propose to be the fundamental building blocks of complex spatio-temporal neural activity patterns.
By applying one simple equation that governs various natural phenomena — the eigendecomposition of the Laplacian — to the structure of the human brain, we predict the building blocks of the brain’s functional patterns. In this way connectome harmonics provide a mapping from the brain’s neuroanatomical structure (the connectome) to its function (the electromagnetic wave patterns that the brain produces).
What is self-organization in the brain context?
Self-organisation in the context of the brain can have multiple dimensions. For instance, the brain’s structural self-organisation is a well-studied area in neuroscience. Another dimension would be the functional self-organisation in the brain, which focuses on how the meso-scale or macro-scale patterns of brain activity self-organise or emerge from the interaction of individual neurons.
My work with the connectome-harmonic framework mainly focuses on understanding the principles underlying this functional self-organisation in the brain. In our brains we have different types of neurons, yet they can be categorised into two main categories: excitatory neurons, whose activity leads to excitation of other neurons; and inhibitory neurons, which inhibit the activity of other neurons.
Interestingly, the interplay between these two opposite types of activity turns out to be the very mechanism that is needed for the self-organziation of structured patterns [in the brain]. In his seminal paper[1], Turing introduced a possible explanation for the biological pattern formation and morphogenesis based on the very same mechanism of activator-inhibitor dynamics, which created a new field of study using theoretical models called reaction-diffusion models.
In our work we use a similar type of model, which describes the interactions between excitatory and inhibitory neurons over a neural population in order to understand how these neuronal interactions give rise to global patterns that emerge at the whole brain. In self-organising systems there is no one ‘leader’ that tells what each element should do, but rather the global pattern naturally emerges from the activity and interactions between individual elements.
When we talk about functional self-organization in the brain, we refer to the emergence of the macro-scale patterns from the interactions of individual elements, neurons or neural populations.
What does your work on connectome-specific harmonic waves tell us about the nature of consciousness?
My work on connectome-specific harmonic waves aims at extracting characteristic signatures of different states of consciousness. So instead of directly tackling the hard problem of consciousness or trying to gain insight into the nature of consciousness, I try to start with a simpler problem and focus on understanding the neural [harmonic] correlates of different states of consciousness such as wakefulness, sleep, psychedelic state, meditation, anaesthesia and disorders of consciousness like vegetative or minimally conscious states.
So I take the functional neuroimaging datasets such as functional magnetic resonance imaging (fMRI) data acquired in each of these different states of consciousness and translate it into the new language that connectome harmonics define.
In a way it is similar to taking a musical piece and decomposing it into its musical notes. The connectome harmonics can be seen as the building blocks, the musical notes, that together compose a complex musical symphony — our brain activity.
When we apply this connectome-harmonic decomposition to the fMRI data acquired in various different states of consciousness, we see how the brain is actually playing a different kind of music in each of these different states. For example, in deep sleep or anesthesia the brain activity is mainly composed of low frequency connectome harmonics, whereas in the psychedelic state we see the opposite effect.
So the connectome harmonic framework aims at linking the brain activity to the experienced state of consciousness and reveals that there are frequency-specific signatures that are characteristic to each of these altered states of consciousness.
In terms of the characteristic signatures of different states of consciousness you mention, are humans mostly the same in terms of these signatures being observed in the various states you mention? Or is there a lot of diversity between individuals?
Clearly there are individual differences when we consider the brain activity between two people in the same state, but when we talk about connectome harmonic signatures of these states, we talk about general trends that we see in the whole group. It’s a bit like the fact that in nature no two zebra patterns are exactly the same but all zebra patterns share certain key characteristics that allow us to classify them immediately. Each of these brain states shows certain characteristics that we refer to as the signatures of that particular state, such as the loss of energy at a certain part of the connectome harmonic spectrum and increase in energy in another part or the size of the repertoire of active connectome harmonics.
Your 2016 Nature Communications paper cites Buzsaki et al. for the notion that “Human brain networks function in connectome-specific harmonic waves,” and that a “characteristic feature of cortical dynamics in mammals is the emergence of behaviour-dependent oscillatory networks spanning five orders of magnitude in the frequency domain.” You and your co-authors then argue, equally remarkably, that “a ubiquitous mathematical framework, eigendecomposition of the Laplace operator, which lies at the heart of theories of heat, light, sound, electricity, magnetism, gravitation and fluid mechanics, can predict the collective dynamics of human cortical activity at the macroscopic scale.” What kinds of predictions are you referring to here? Has your work since this paper was published supported your contention?
What we are referring to is that known brain activity patterns measured in resting state such as the resting state networks, can be purely predicted from the structure of the human brain simply by applying this ubiquitous mathematical framework. Empirically, we also demonstrate how the connectome harmonic patterns that we estimate using only the structure of the brain, actually match the brain activity patterns in the resting state.
In our 2016 paper, we also show that connectome harmonics are the building blocks of pattern formation that result from mathematical models describing brain activity, such as those known as neural field models. These models explain how global (macroscopic) patterns of brain activity can emerge from the interaction between neurons or neural populations and we show that these macroscopic patterns, this collective activity of neural populations, is in fact a combination of connectome harmonics.
Yet there is another very important characteristic of these harmonics: they actually provide a “function basis” to describe any pattern of brain activity. After we published our first paper on connectome harmonics in 2016, we started focusing on this property and used the set of all connectome harmonics as a new harmonic language, to describe brain activity in different states of consciousness. Our later studies reveal that differences that remain hidden in the standard representation of the functional brain data become revealed when we describe the data in this new harmonic language. This efficiency provides further support to our hypothesis that the brain may be operating through these harmonic patterns, and different states that we experience simply match the activation and deactivation of these harmonic building blocks of brain activity.
Do you consider the connectome harmonics to be causal or only consequences of the underlying neuroanatomy and not themselves causal? That is, are the harmonics more like the epiphenomenal train whistle or they actually a part of the causal machinery of our consciousness?
Based on our findings, I consider them to be a very effective mechanism that nature has chosen to create/represent our experience. What our findings say so far is; when our experience — the state of consciousness — changes, so does the composition of brain activity in terms of connectome harmonics. We can’t make any conclusion about them being the causal machinery of our consciousness, as we only study the neural correlates of consciousness at this stage. Yet intuitively, I would say that they are like the sound waves that emerge when we play a musical note in a musical instrument, and mathematically they are the equivalent of these waves adapted to the structure of the human brain. The same way the combination of these sound waves gives rise to what we experience as music, a particular combination of connectome harmonics gives rise to a pattern of brain activity that we experience as a particular state of consciousness.
How could we definitively answer the question of electromagnetic fields being causal or not? Could we look to the correspondence between subjective states and these field harmonics, then compare subjective states to synaptic dynamics, and assess whether synaptic dynamics have the spatial and temporal granularity required to explain changes in observed subjective states? Buzsaki, as long ago as 2004 suggested that electromagnetic oscillations are probably causal (“These oscillations are phylogenetically preserved, suggesting that they are functionally relevant. Recent findings indicate that network oscillations bias input selection, temporally link neurons into assemblies, and facilitate synaptic plasticity, mechanisms that cooperatively support temporal representation and long-term consolidation of information,” Buzsaki (2004) “Neural oscillations in cortical networks”), so this has been quite a long debate in the field, as you know!
I think it may be important to distinguish between ‘causal’, ‘functionally relevant’ and ‘epiphenomenal’. My whole research so far provides evidence for the functional relevance of whole brain oscillations and emerging standing waves for the state of consciousness and even for human cognition.
In that respect, I would think it is very unlikely for them to be epiphenomenal. So far, we have looked at how the energy of connectome harmonics changes in various different states of consciousness including altered states of consciousness induced by LSD, psilocybin and ketamine; propofol-induced anesthesia, disorders of consciousness (DOC) with responsive and non-responsive patients; meditation; as well as sleep. What is very robustly visible in all of these studies is the energy changes across the connectome harmonic spectrum, which is unique to the state of consciousness. For instance, the psychedelic state induced by LSD and psilocybin show very similar connectome harmonic energy signatures and this signature is the exact opposite of what we find in propofol-induced anesthesia and non-responsive DOC patients.
This again indicates the functional relevance of these harmonic brain states for consciousness. We also conducted some computational modeling studies (Atasoy 2016, and a novel paper that is under review at the moment) to show the mechanistic relationship between these oscillations changing together with changes in the excitation/inhibition balance in the brain, which in turn affects the state of consciousness.
Lastly, in a recent study we also looked at the functional relevance of harmonic patterns for human cognition and found clear evidence for the role they play not only in our cognition but also even in the structure of the cortical organization; in the emergence of how functionally specialized areas in the cortex.
So based on all these studies, I would conclude that they are at least functionally relevant rather than epiphenomenal. It is more difficult to talk about their causality for consciousness. Even though the link between connectome harmonic energy changes observed in pharmacologically-altered states of consciousness such as psychedelics and anesthesia can be interpreted as evidence for their causal role for consciousness, I would not reach that conclusion at this stage. I think to conclude a causal relation between any neural correlate and consciousness, we need to have a clearer understanding of the nature of the link between brain and consciousness first, which is what we are working towards as the neuroscience community.
How does your work compare to that of Wolfgang Klimesch or Pascal Fries, both of whom have also developed taxonomies and theories about electromagnetic field dynamics (Klimesch’s binary hierarchy brain body oscillation theory, Klimesch 2018, and Fries’ Communication Through Coherence model, Fries 2005, 2015)? For example, Klimesch makes explicit a nested binary hierarchy relationship between various frequency bands of brain oscillations as well as body oscillations (e.g., 1:2 relationships between theta and alpha, beta and gamma, etc., center frequencies).
I would say our work supports these findings and theories; and complements these ideas by bringing in the spatial component and its relation to these temporal oscillations. By applying a fundamental principle to the structure of the human brain, we estimate the particular shapes of harmonic patterns that would emerge (spatially) in the brain from the synchronization of these nested temporal oscillations as studied in Wolfgang Klimesch’s work. Our findings also support Pascal Fries’ Communication Through Coherence model, as we reveal various different modes of communication through synchronous oscillations in the cortex in terms of harmonic patterns emerging on the human connectome (see also Glomb, K. and Kringelbach, M. L. and Deco, G. and Hagmann, P., Pearson, J. and Atasoy, S., Functional harmonics reveal multi-dimensional basis functions underlying cortical organization, bioRxiv, doi: 10.1101/699678)
Beyond the very poignant pleasures of scientific discovery and understanding brain and consciousness dynamics, what practical impacts do you see resulting from your work in connectome harmonics? Might we see new therapeutics for sleep disorders, dementia, depression, etc.?
I hope so. My long-term vision is to take this fundamental understanding that we are gaining about the brain and apply it for both diagnostic and therapeutic purposes. It is like first learning the language of the brain, and then utilizing this language to support the brain’s natural processes, for our health and well-being.
[1] Turing, A. M. (1990). The chemical basis of morphogenesis. Bulletin of mathematical biology, 52(1–2), 153–197.
