Connectome-based models of whole-brain dynamics: from theoretical principle to practical application

Davide Momi Organizer
Center for Addiction and Mental Health
Toronto, Ontario 
Joana Cabral Co Organizer
University of Minho
University of Minho
Braga, Portugal 
John Griffiths, PhD Co Organizer
University of Toronto
Toronto, Ontario 
Sunday, Jun 23: 9:00 AM - 5:30 PM
Educational Course - Full Day (8 hours) 
Room: Grand Ballroom 102 
The topic of whole-brain modeling stands at the forefront of contemporary neuroscience, gaining remarkable traction within the OHBM community. Evidenced by the exponential surge in publications, from fewer than 50 in 2013 to nearly 600 by the close of 2023, this surge signals a pivotal moment for our field. It reflects a collective thirst for deeper understanding and technical prowess in unraveling the complexities of brain organization. Recognizing this momentum, we propose an Educational Course, a continuum of our prior successful endeavor at OHBM. In the previous edition, our course garnered immense acclaim; attendees filled the room, captivated by the engaging sessions.
Our plan for the workshop structure is to have three sections: Understanding the Fundamentals: Theory and Background, In Action: Cognitive Systems and Applications, and Beyond Theory: Clinical Applications and Practice. Each section would have 2-3 tutorial lectures, and importantly also conclude with a ~1hr hands-on session, where participants spend time working through pre-prepared tutorial code, pertaining to one or more of the preceding lectures.
An intrinsic aspect of our course's success lies in the hands-on sessions, where attendees actively engage with tutorial code. These sessions have not only captivated the audience, as evidenced by the overflowing room, but have also laid the foundation for an invaluable resource. The materials and content from these sessions are publicly accessible (, fostering widespread learning. Our commitment to enhancing accessibility and knowledge dissemination is further evident through our ongoing collaboration with Neurolibre, where we are diligently working to publish the course notebook.
Crucially, during our previous edition of the course, we actively collected after each hands-on session. This iterative approach ensures our course remains dynamic, continuously improving to meet the evolving needs and expectations of our audience. As we stand on the precipice of a new OHBM edition, we are confident that this Educational Course will not only meet but exceed the high expectations set by its predecessor. Its timeliness, evidenced by the flourishing interest in whole-brain modeling, coupled with its proven success and commitment to improvement, renders it an indispensable addition to the OHBM educational repertoire.

At the end of this Educational course, attendees will understand:
- What the conceptual ingredients involved in modelling brain network dynamics at the macro-scale are, and why these tools are important in contemporary neuroscience
- How to construct and simulate activity in a brain network model using i) an anatomical connectome or other arbitrary network, and ii) a given choice of neural dynamics
- What the typical effects are of applying spatially and temporally patterned stimulation to a brain network model
- The effect of brain stimulation on whole-brain network dynamics
- How psychiatric changes can be represented in whole-brain models, and what additional information can be extrapolated from such simulations
- What aspects of whole-brain models, and the neuroimaging data they are built from, lend themselves to development of personalized simulations

Please find the full education course Agenda below:

08:00-08:40 ~~ MORNING COFFEE & GREETS ~~

08:40-09:00 - Introduction to the workshop - JD Griffiths
09:00-09:40 - Introduction to Dynamics - V Jirsa
09:40-10:20 - Criticality, metastability, multistability: The "primitives" of complex brain dynamics - M Breakspear
10:20-11:00 - Intro to connectome-based neural mass modeling - SP Bastiaens
11:00-11:20 - Panel discussion
11:20-12:00 - Hands-on Session 1 - D Momi/JD Griffiths

12:00-12:45 ~~ LUNCH ~~

2) In Action: Cognitive Systems and Applications

12:45-13:25 - Waves or Networks at the basis of Cognition? - J Cabral
13:25-14:05 - Brain geometry and dynamics - JC Pang
14:05-14:45 - Modelling of brain stimulation and network dynamics - D Momi
14:45-15:05 - Panel discussion
15:05-15:45 - Hands-on Session 2 - D Momi/JD Griffiths

15:45-16:00 ~~ BREAK ~~

3) Beyond Theory: Clinical Applications and Practice

16:00-16:30 - Psychedelic Drugs and Brain Dynamics - J Cruzat
16:30-17:10 - Neurocomputational modelling for predicting psychosis - A Diaconescu
17:10-17:20 - Panel discussion
17:20-17:30 - Hands-on Session 3 - D Momi/JD Griffiths


1) Linear and Nonlinear Brain Network Dynamics: the course will provide a broad introduction to the field of nonlinear dynamics, focusing both on the mathematics and the computational tools that are so important in the study of chaotic systems like the brain. The students will learn about complex system, brain dynamics, chaos, fractals, information theory, self-organization, agent-based modeling, and functional brain networks

2) Personalized Brain Network Modeling: theoretical background of large-scale brain network modeling, simulation of resting-state networks, brain disorders, concepts of nonlinear dynamics (bifurcation analysis, phase plane, manifolds, etc.), parameter optimization and model inference, application of brain network modeling for clinical questions, visualization multimodal brain dynamics

3) Clinical Applications of computational modeling: how to use whole-brain computational model to predict and potentially explain brain psychiatric and neurological disorders  

Target Audience

Our target audience is a broad and multi-disciplinary slice of the OHBM community, including i) cognitive / clinical neuroscientists trained in biological, psychological, and medical sciences, who are fairly naive on the details but recognize the value of the approach; ii) physicists / engineers / mathematicians who have had little prior exposure to multimodal neuroimaging and brain network mapping, and iii) individuals with some familiarity with whole-brain modelling already, and are looking to extend and consolidate their extant technical knowledge. The breadth of topics covered in the workshop make it suitable for people with varying levels of experience.  


1. Introduction to Dynamics

This introduction provides a concise overview of dynamic systems in neuroscience, emphasizing their significance and practical applications. It covers fundamental concepts and tools, examining dynamic behaviors across various degrees of freedom with examples from computational neuroscience. The course explores intricate mechanisms like criticality, bifurcations, and symmetry breaking, highlighting their roles in shaping complex behaviors in neuronal networks. It addresses self-organization phenomena within these networks, shedding light on the organizational principles governing neural systems. Additionally, the course makes the connection to probability density distributions, Free Energy, and the Maximum Information Principle. It elucidates the interplay of deterministic and stochastic forces, offering a nuanced understanding of their impact on neural dynamics.
In summary, this lecture delivers a comprehensive introduction to dynamical systems in neuroscience, enhancing our understanding of neural phenomena. It shall serve as a valuable resource for scholars, researchers and students studying the mechanisms of neural networks. 


Viktor Jirsa, Institut de Neurosciences des Systèmes
Aix Marseille University
Marseille, N/A 

2. Criticality, metastability, multistability: The "primitives" of complex brain dynamics

Considerable research suggests that multi-scale processes in the brain arise from so-called critical phenomena that occur broadly in nature. Criticality occurs in systems perched between order and disorder, allowing agents to quickly adapt to a dynamic environment. But criticality comes in several flavours that possess unique computational properties, namely bifurcations, metastability and multistability. Each of these complex dynamics builds on simple underlying dynamical processes in different ways to broaden the behavioural repertoire of dynamical systems. I will illustrate these distinct types of criticality in simple neural mass models and overview methods to detect and disambiguate them in functional neuroimaging data. I will also summarise their explanatory potential in brain health and disease, ranging from natural vision, motor behaviour, decision making, seizures and hallucinations  


Michael Breakspear, University of Newcastle Newcastle, New South Wales 

3. Introduction to connectome-based neural mass modelling

Mathematical models of human brain activity have been central in gaining insights into the hidden mechanisms of the underlying neural processes at multiple scales. In this context, Whole-Brain Modelling (WBM) is a sub-field of computational neuroscience concerned with building comprehensive theoretical and computational models that represent and simulate the neural activity across the entire brain. The common objective of this approach is to investigate the mechanisms through which macroscopic spatiotemporal patterns of neural activity can be explained by studying the interplay among anatomical connectivity structure, intrinsic neural dynamics, and external perturbations (sensory, cognitive, pharmacological, electromagnetic, etc). Such macroscopic phenomena (i.e brain oscillations), and models thereof, are of particular scientific interest because a) large scale neural activity can be most readily obtained from the brains of healthy human subjects, using noninvasive neuroimaging and related methods b) they represent neural systems in a holistic and relatively intact state, . Simulations of human brain activity, in both health and disease, are therefore a principal focus of current WBM research.
The overarching idea is to model the brain at the macroscale as a network of interconnected regions, which are defined by (principally) neuroimaging-based brain parcellations The presence and weights of the network edges interconnecting the nodes are then derived from neuroimaging- or chemical tract tracing-based anatomical connectivity measurements. The nodes can be described using neural mass models (NMM) which represent the coarse grained activity of large populations of neurons and synapses using a small number of equations to express their mean firing rates and mean membrane potentials. NMM are capable of describing the change in firing rate of neural populations without spatial information and spatiotemporal time delays providing a succinct yet biophysically meaningful description of brain activity at the mesoscopic scale to reflect phenomena observed empirically at the macroscale. The main advantage of NMM is that the simplification of the dynamics reduces the number of dimensions or differential equations that need to be integrated enabling us to hone in on the behavior of a large number of ensembles and understand more clearly their dynamics. The aim of those models is to propose a balanced model between mathematical tractability and biological plausibility while still reproducing a wide range of empirical data features across multiple measurement modalities.
These features include: fast oscillations in local field potential (LFP) and extracranial electromagnetic (MEG, EEG) signals; slow quasi-periodic activity fluctuations in haemodynamic (BOLD fMRI, fNIRS) signals; inter-regional synchrony/covariance (‘functional connectivity’) and causal interactions (‘effective connectivity’) in both fast and slow activity patterns; sensory- or electromagnetic stimulation-evoked response waveforms; graph-theoretic properties large-scale network activity; and many others. 


Sorenza Bastiaens, Unversity of Toronto-CAMH Toronto, N/A 

4. Waves or Networks at the basis of Cognition?

How the cognition ‘emerges’ from the physical structure of the brain remains unclear. Neuroimaging studies reveal signatures of cognition in the dynamics of large-scale functional networks, whose origin and generative mechanisms remain under debate. Efforts have been made to link the formation of functional networks at the macroscale to neuronal activity at the microscale using whole-brain computational models. While these models have served to support distinct mechanistic hypothesis for the genesis of functional networks, an alternative hypothesis relating functional networks to resonance phenomena is emerging. In my talk I will discuss these two distinct perspectives, reinforcing the importance to maintain openness to different mechanistic hypothesis while new evidence is needed to disambiguate current conflicts. 


Joana Cabral, University of Minho
University of Minho
Braga, Portugal 

5. Brain geometry and dynamics

The dynamics of many physical systems are naturally constrained by their underlying structure. Here, I will show that the nervous system is no exception, with geometric eigenmodes derived from the brain’s cortical and subcortical geometry accurately capturing diverse experimental human functional magnetic resonance imaging (fMRI) data from spontaneous and task-evoked recordings. Moreover, these geometric constraints are unique to each individual and universally exist across different species. Finally, I will show that the close link between geometry and function is explained by a dominant role of wave-like activity, and that wave dynamics can reproduce numerous canonical features of functional brain organization. These findings identify a previously underappreciated role of geometry in shaping function, as predicted by a unifying and physically principled model of brain-wide dynamics. 


James Pang, PhD, Monash University Melbourne, Victoria 

6. Modelling of brain stimulation to unveil signal propagation and network dynamics

The human brain comprises distinct resting-state networks (RSNs) characterized by spontaneous activity patterns. Despite this highly structured functional pattern, its laws of motion and principles of organization have proven challenging to understand with currently available measurement techniques. In such epistemic circumstances, an extremely elegant modus operandi to investigate brain complexity with high spatial and temporal resolution entails the administration of precise and synchronized external stimulation, followed by a meticulous examination of the resulting induced propagation dynamics that emerge in response to these perturbations. In this framework, a combination of empirical stimulus-evoked data analyses and whole-brain, connectome-based neurophysiological modelling provide an elegant scaffold to investigate questions around the physiological basis and spatiotemporal network dynamics of RSNs’ activity. Deciphering this evoked propagation pattern is essential for a comprehensive understanding of the brain's response to stimulation and therefore for personalized and targeted interventions, with potential applications ranging from therapeutic treatments to cognitive enhancement. 


Davide Momi, Center for Addiction and Mental Health Toronto, Ontario 

7. Psychedelic Drugs and Brain Dynamics: Unveiling Turbulent Signatures and Control Energy Landscape Flattening

Psychedelics like LSD and psilocybin alter subjective experience through serotonin 2A (5-HT2A) receptor agonism, resulting in increased brain entropy. We propose this heightened entropy reflects a flattening of the brain's control energy landscape. Using fMRI data, we show that LSD and psilocybin reduce control energy for brain state transitions, leading to more state changes and increased entropy. Analysis tying 5-HT2A receptor distribution to control energy supports this link. Our study reveals how psychedelics facilitate state transitions and diverse brain activity and demonstrates the potential of receptor-informed network control theory. Additionally, psychedelics show promise as treatments for neuropsychiatric disorders. We explored how LSD and psilocybin impact the brain's functional hierarchy using a novel turbulence framework. Both psychedelics produced distinct turbulence-based changes, affecting higher-level networks, especially the default mode network. These findings support the hypothesis that psychedelics modulate the brain's functional hierarchy and offer quantification for two different psychedelics, with potential implications for therapy.  


Josefina Cruzat, Universidad Adolfo Ibanez
Santiago, N/A 

8. Neurocomputational modelling for predicting outcomes and characterizing neurocognitive pathophysiology in youth at clinical high risk for psychosis

Schizophrenia is a debilitating psychiatric disorder that imposes significant socio-economic burdens on individuals and society. Thus, early identification of those at "clinical high risk" (CHR) of developing schizophrenia is imperative. While interventions at this stage can potentially thwart the onset of schizophrenia or related psychotic disorders, even those CHR patients who do not go on to develop schizophrenia frequently continue to experience high symptom burden and functional impairment. Thus, there is a compelling need to elucidate additional prognostic indicators for this cohort to prioritize treatments.

Rationale and Study Aims: A promising avenue is employing neurophysiological measures, specifically, event-related brain potentials (ERPs). Two particular ERPs, the mismatch negativity (MMN) and the N400 semantic priming effect, have been shown to predict conversion to psychosis (1) and decline in psychosocial functioning 1-year later (2), respectively in CHR individuals. Despite the potential of these ERP biomarkers, their neural mechanisms remain largely unknown, thus limiting their clinical utility.

Methods: In this study, we address this by fitting a connectome-based neural mass model (CNMM) (3) to both auditory MMN and N400 datasets in a group of N=47 CHR individuals, whose symptoms and general social and role functioning was assessed at baseline and 1-year later (4,5). In this CNMM model, neural dynamics at each source are described by Jansen-Rit (JR) equations (6,7), which encapsulate neural dynamics across three populations: pyramidal neurons, excitatory, and inhibitory interneurons, forming a circuit with one positive and one negative feedback loop. After fitting the CNMM model to participants’ grand averaged ERPs across the two datasets, we extracted the local gain parameters (C1-C4) and used them to predict changes in symptoms and psychosocial functioning.

Results: In the MMN task, we found that increased excitation via increased excitatory-to-pyramidal connectivity, was observed in the primary auditory, middle cingulate, and inferior frontal areas, and was associated with an increase in positive symptoms one year later (r= 0.53, p = 0.019). In the N400 priming dataset, heightened disinhibition characterized by a decrease in inhibitory-to-pyramidal and an increase in pyramidal-to-excitatory connectivity—across the occipital, precuneus, middle cingulate, and inferior frontal regions was linked to diminished social and role functioning after one year (r= 0.42, p = 0.031; r= 0.48, p = 0.012), respectively. These results provide support for a neurophysiological model of the psychosis prodrome, linking excitatory-inhibitory mechanisms, brain connectivity, clinical symptoms, and long-term functional outcomes.


1. Hamilton HK, Roach BJ, Bachman PM, Belger A, Carrión RE, Duncan E, et al. Mismatch Negativity in Response to Auditory Deviance and Risk for Future Psychosis in Youth at Clinical High Risk for Psychosis. JAMA Psychiatry. 2022 Aug 1;79(8):780–9.
2. Lepock JR, Ahmed S, Mizrahi R, Gerritsen CJ, Maheandiran M, Drvaric L, et al. Relationships between cognitive event-related brain potential measures in patients at clinical high risk for psychosis. Schizophr Res. 2020 Dec;226:84–94.
3. Momi D, Wang Z, Griffiths JD. TMS-evoked responses are driven by recurrent large-scale network dynamics. eLife. 2023 Apr 21;12:e83232.
4. Charlton CE, Lepock JR, Hauke DJ, Mizrahi R, Kiang M, Diaconescu AO. Atypical prediction error learning is associated with prodromal symptoms in individuals at clinical high risk for psychosis. Schizophrenia. 2022 Nov 25;8(1):1–10.
5. Lepock JR, Sanches M, Ahmed S, Gerritsen CJ, Korostil M, Mizrahi R, et al. N400 event-related brain potential index of semantic processing and two-year clinical outcomes in persons at high risk for psychosis: A longitudinal study. Eur J Neurosci. 2023 Jun 29;
6. David O, Friston KJ. A neural mass model for MEG/EEG: coupling and neuronal dynamics. NeuroImage. 2003;20:1743–55.
7. Moran RJ, Kiebel SJ, Stephan KE, Reilly RB, Daunizeau J, Friston KJ. A neural mass model of spectral responses in electrophysiology. NeuroImage. 2007;37:706–20.  


Andreea Diaconescu, University of Toronto Toronto, Ontario