Hello. Good morning. My name is Vinod Menon. I'm Professor of Psychiatry, Neurology and Neuroscience at Stanford University. And today I'm going to be presenting some ideas, both theoretical and applied, on the topic of brain networks and psychiatric disorders. And the goal of this research overall is to advance foundational knowledge of human brain function and dysfunction, and to use this knowledge to help children and adults with psychiatric disorders. And the approach is decidedly a cognitive and systems neuroscience based framework for understanding psychiatric disorders, as tremendous advances have been made in the past two decades in understanding how the human brain processes information. And the goal is to try and bring that knowledge with a deep understanding of the theoretical framing of systems neuroscience to understand psychiatric disorders in a deeper manner than has been previously possible. So the outline of this presentation is as follows. I'm going to talk about why brain networks? What are the basic principles? How is the brain intrinsically organized in the neurotypical individual? And a big focus of the presentation today is on cognitive control networks. And these are networks that flexibly and adaptively allocate mental resources to permit the dynamic selection of stimuli, thoughts, and actions in response to context-specific goals and intentions. And we'll see that there is a canonical set of cognitive control networks that are very typically activated during and consistently activating during a range of cognitive tasks involving potential control, inhibitory control, set shifting, working memory. And this has led to the notion that there are core cognitive networks involved in cognitive information processing. And from the viewpoint that most, if not all, psychiatric disorders involve at some level or the other impairments in cognitive control, the study of these core networks has become a central focal point for trans-diagnostic as well as studies of individual psychiatric disorders. So at the core of trying to understand the role of brain networks in psychiatric disorders is the notion that I alluded to of cognitive control networks that are impaired in psychopathology. And so I'm going to talk about what these networks are, what regulates control of these networks, how do they switch between one being active and the other not being active so as to allocate resources that are necessary for the task at hand. And it's really impairments in these core processes that are foundational to understanding cognitive inflexibility and internal mental thought processes that go awry in psychiatric disorders. And in particular, I'm going to focus on a triple network model of psychopathology involving three core networks, the salience, frontal parietal, and default mode networks, which I'll describe at some length so you have a better understanding of what they actually do in neurotypical cognition and how impairments in these networks and how their interactions give rise to a range of phenotypic features we see in psychiatric disorders. So once I develop this general framing from the basics to an understanding of cognitive control networks and their role in psychopathology broadly, I'm going to provide a number of examples from our own and other work that focuses on brain network dysfunction in autism. So this is kind of going from the developmental span to ADHD, schizophrenia, and depression. And there's some unifying themes that come across all of these dissumers, and that's the focus that I want to emphasize here. And so while we can actually do a lot of things, we can construct examples of brain networks, the goal is to try and understand system dysfunction at a very fundamental basic level. So there are some unifying themes, both in terms of trying diagnostic aspects of psychiatric disorders, but also in terms of cognitive and neural systems, neuroscience work on how the cognitive control systems function in the brain. So I want to emphasize that there are historically, there's, you know, the way to study brain function from a neuropsychological point of view was through lesion studies. So here's kind of the famous example of Phineas Gage, who was a railroad worker, and he was quite a normal person before this incident happened, whereas a spike went through his medial prefrontal cortex, and it was then noted that he was not the same person anymore. He had become antisocial, bad manners, could not hold a job, and grossly profane. In other words, no longer human, not being able to interact in a socially meaningful manner, engaging them with others as a result of this injury. So now, these kinds of lesion studies have been the mainstay of, for, I would say, the 19th century for trying to understand what happens in the early part of the 20th century, in terms of trying to understand what happens, what serious impairments occur as a result of brain damage. Now, this is not a very good model for studying psychiatric disorders, because we almost never see gross lesions. Individual brains do not have the same architecture. The structural abnormalities are somewhat subtle. And this has given rise to the idea that deficits in psychiatric disorders primarily arise from damage to brain circuits and networks, as I'll describe. And there is also this notion that it's not just the damage to the current system, but there's a whole history of how the individual from childhood has experienced various events, whether they are stressful, or are traumatic, or they have particular deficits in signaling individual parts of the brain, which are dysfunctional as a result of imbalance between excitatory and inhibitory neurons. All of those things can have an impact on how an individual at a particular age, when we study them, process information. So now the study of brain networks has a core anchor in the anatomical framing of Brodmann. And this is our cytoarchitectonic maps that you might be very familiar with. This already gives rise to the notion that individual brain areas have different local circuit properties, the way that the neurons are expressed across the layers, the number of layers, the cell types, and so on. So each region has a slightly different neuronal footprint compared to other regions. And so this is one of the anchors from which we think and study brain circuits and brain anatomy and brain networks in humans. Now, with the advent of modern neuroimaging tools, such as functional brain imaging, with its advent in the early 90s, the shift, there's been a shift in terms of trying to understand how the functional architecture, not just where the structural deficits might be gross or subtle, but in terms of understanding the overall functional organization. And several tools have been developed and they've become the mainstay of trying to understand how the human brain is functionally organized. And this forms the basis for trying to understand both normative information processing in brain circuits, but also dysfunction and psychiatric disorders. So this is a movie showing snapshots every half a second of how the local activity in the brain is changing. And you can see this is a highly dynamic complex system. And so to be able to understand this in a deep and fundamental way, we have to first gain an understanding of the overall organizing principle of how this is structured. Is it really a fully random set of circuits and nodes, which are interacting at will, or is there some underlying principle? And that's kind of the core of trying to understand the functional architecture. And that's the impetus for trying to understand this. So the use of these technologies has brought new ways of constructing and understanding brain circuits and networks, but there's also historical antecedents here as illustrated from the work of Marcel Messelin based on lesion studies or studies in macaque non-human primate models of human cognition. And some of these are the spatial attention network, which is anchored in the posterior parietal cortex and frontal eye fields. The language network, of course, which is uniquely human, anchored in Broca's and Bernicke's areas. And a memory system anchored in the hippocampus with strong links to frontal parietal cortex, face recognition system in the inferior temporal cortex and temporal polar cortices, and a working memory system anchored in the dorsolateral prefrontal cortex and parietal cortex. So even before the advent of modern imaging techniques, there was this general sense that there are dedicated systems for processing certain kinds of information so that if you're processing language, the system would be more likely to be engaged than a spatial attention system and so on. So this gives us an anchoring for thinking about, you know, how we might demarcate brain networks, and they all have specific sets of nodes and links within the architecture that I laid out from Brodman. So what is a brain network? And I want to spend some time thinking about discussing the general fundamental features, because I think they're really important, both from a theoretical point of view, and also from the point of view of understanding individual components of systems that are dysfunctional in various psychiatric disorders, because this has to be done in some principled and thought out manner. And so a brain network is a set of coherent nodes, which are linked together. And here's an example for network one and network two. These are not completely independent, but they mostly reside in ways that if one node of this network is active, these nodes are also very likely to be active. So for example, when you're doing a working memory task, and these networks are constantly interacting with other systems, but those are more sporadic and context and need based. And so the notion that each system is very tightly coupled, but it also has the ability to interact dynamically with other systems. And this is really what constitutes the core of trying to understand why we should think about brain organization in terms of functional networks. And this is again, in relation to the slide I presented earlier, showing individual specialized functional systems for cognition. So the general notion is that multiple brain areas constitute a network in the sense of being consistently co-active across a range of tasks. They serve dedicated functions, but their functions are also, should be contextualized in the context of their influences on other networks, whether they be to activate them or suppress them in the service of goal-directed behavior. And individual nodes of the networks are difficult to destabilize and typically function in concert. It's not that the nodes have identical functions because each node, even within a network, receives different types of inputs and sends different types of outputs, but they're generally difficult to destabilize. And so that's why when we do functional neuroimaging studies of tasks, such as working memory, you'll see almost always co-activation of the dorsolateral prefrontal cortex and the posterior parietal cortex. So one other way, principle that we really should think about as we try to figure out the notion of how the brain is, the human brain is functionally organized, is the notion that each area, as I referred to when I presented the Broadman map, has a certain intrinsic property arising from cellular distribution, the layers, which vary across regions and so on, and its extrinsic connections. So this is what gives rise to the functional properties of a cortical area. And moreover, each cortical area has a unique connectional fingerprint, and these fingerprints distinguish the function of one cortical area from another. So even if they have the same identical cytoarctic tectonic properties, the functions may be different because they're connected to different parts of the brain so that the inputs and outputs are different. And here's kind of a toy example that kind of shows that each part of the brain, for example, in this cartoon diagram, is connected differently. So the areas that this node is connected to, or this part of the brain is connected to, is different from this. And moreover, you have this notion that you don't have a random architecture, but you have hubs which help to interconnect different parts of the system. And this notion is also really critical in trying to understand how networks function. And this is a polar plot showing two different areas in the prefrontal cortex, showing that they have different patterns of connection. So even neighboring areas can have very different connections and therefore functions. And this is just a broad expansion of that, showing a range of nodes and their interconnections. So this is a very complex architecture. We saw the movie which highlighted this kind of dynamic connection. And the goal is to try and understand the key aspects of the functional architecture of the system, so that we can then bring it to bear on both the study of cognitive function and affective function, as well as psychiatric, their disruption in psychiatric disorders. So how do we go about constructing and trying to understand how networks are organized? So there are multiple computational tools that have been brought to bear on this. And I want to highlight three general methods that are used to construct brain circuits or networks. And so if you took something like this resting state fMRI and looked at connectivity of two neighboring areas, you can see that these two neighboring areas in the inferior frontal gyrus and the anterior insula have different patterns of connectivity with respect to the lateral frontal parietal and temporal areas. And you can see this anterior insula with very strong links to the anterior cingulate cortex. You can see the inferior frontal gyrus with more strong links to the dorsal lateral prefrontal cortex. This is already suggesting and illustrating a general principle that even neighboring areas can have, brain areas can have very different patterns of connectivity and their functions therefore depend on this overall pattern. The other approach that has been used is using a functional connectome kind of point of view where you see different brain areas. For example, the Brodman map and then you construct a functional connectivity matrix and then you use a modularity detection algorithm to detect what are the core nodes of each network or community and what are the nodes that actually link multiple networks. And then you can study them in both in the context of how they're intrinsically organized, how they get perturbed by specific cognitive tasks, and also how their organization might be impaired. The other approach is to use something called independent component analysis, ICA decomposition. And there are similarities and, you know, differences between these methods, but they all converge on very similar kinds of functional systems, at least in the context of the present talk in relation to psychiatric disorders. And core among which, and I'll kind of emphasize more on some of these systems, is a salience network, which is anchored in the anterior insula and the anterior cingulate cortex, the so-called central executive or lateral frontal parietal network, both in the left and right hemispheres, with the dorsolateral prefrontal cortex and the posterior parietal cortex, anchored in the supramarginal gyrus, and then the default mode network, which is anchored in the posterior cingulate cortex and the ventromedial prefrontal cortex. So these are segregated systems that subserve distinct functions. They interact, as I mentioned, but also they have this unique network-like feature in the sense that if a node of this network is active, for example, if the anterior insula is active, it's almost always the case that the anterior cingulate cortex is active as well. And the same goes for other systems as well. And so how do these systems function coherently? How do they interact with other systems? That's been the core of a lot of work on trying to figure out how they function during cognitive control in a variety of tasks that I mentioned earlier, ranging from the stop signal to the flanker and the working memory tasks. And it's, the main point to emphasize is that these systems have to be studied not only in the context of their intrinsic organization, what's, you know, how is it connected, but also in terms of the functions that they subserve, what kinds of functions, cognitive functions activate what systems and when. And so this has given rise to the general idea that the human brain is intrinsically organized, and this is kind of one of the evolving principles into coherent functional networks with brain areas that are commonly engaged during cognitive task forming systems that can be readily identified intrinsically, and also obviously during the cognitive tasks themselves. So here's kind of an example of some of these functional systems that I alluded to, the frontal parietal network, the default mode network, the frontal eye field and the parietal eye field for saccadic control, and they generally tessellate the whole brain. And so we can identify these networks and study how they function during cognitive tasks. The second major principle that's both relevant to trying to understand how the human brain is organized in terms of large scale networks and also how impairments in a wide range of psychiatric disorders arises is through these series of studies that have shown that you have consistent deactivation in a set of areas, termed the default mode network, which is a tightly functionally and structurally connected system, important for processing self-referential information and for monitoring the internal mental landscape. And I'll unpack these as I go along, and the system is particularly relevant in the context of studying psychiatric disorders. Just to give you a little bit of a historical background, so there was across a range of cognitive tasks, so ranging from language to working memory, episodic memory, it appeared that several areas were actually suppressed during task performance, and this includes the posterior cingulate cortex and the ventromedial prefrontal cortex. We now, of course, know this as the default mode network, which is somewhat quite intriguing because when you ask a subject to do a working memory task, you don't necessarily expect a very consistent pattern of suppression of responses, but this is at a system wide level, we see this suppressed response. And so this is the first study that actually showed that those areas that I highlighted earlier actually form a network, because if you look at the connectivity of one of these regions, like the posterior cingulate cortex, you can uncover most of this network, in particular the ventromedial prefrontal cortex node of this network, as well as the lateral angular gyrus component of the network. So this gave rise to the notion that there is a common system that has a slightly different function from those that are typically engaged during challenging cognitive tasks. And this, of course, model has been replicated with a number of different computational methods. I'm just kind of summarizing here a range of meta-analytic studies which are reviewed in our article here, cited here. And this, the core nodes of this really are in the posterior medial cortex, the ventromedial prefrontal cortex, and the angular gyrus. And this system has been identified using multiple techniques, not just functional neuroimaging, but also structural imaging studies with diffusion tensor, imaging of white matter pathways, showing this very strong cingulate bundle that links the posterior cingulate cortex to the medial prefrontal cortex, this cingulate bundle. And you can have a range of graph theoretic approaches which have also been used to identify them. Suffice to just mention that independent of methodologies and modalities, this is a system that is very reliably identifiable, both at the group level and in each individual. And this just shows, again, the cingulate bundle linking the posterior cingulate cortex with the ventromedial prefrontal cortex, as well as the hippocampus, which is a core node of this default mode network. And this has had, now, given that these areas are actually relatively suppressed during challenging cognitive tasks, this has given rise to the question of what do these, the system actually do? And so I'm just showing you here results of a number of studies and meta-analyses, which are summarized again here, showing that tasks that engage, that involve social cognition, mentalizing and planning, self-referential, where you're thinking about yourself, you're thinking about yourself in the context of others and reading other people's minds, this system is actually engaged. And so this has given rise to the notion that you have a system for monitoring the internal self, both in the context of its planning of future events or referring back to its autobiographical notion of the self, but also how it relates to other individuals in various contexts. And this is just a dynamic movie showing how the, this network is really not static, its responses wax and wane. You can see the posterior cingulate cortex, the ventromedial prefrontal cortex core nodes. And over time, these are fluctuations over time, they wax and wane as the system interacts with other areas in and outside the network. This is a really important illustration of a general principle that these are not static systems. As I mentioned, the notion of the network is that they're commonly active nodes, but they also interact with other systems. And it turns out, and I'll illustrate this through multiple examples, subveniences in network dynamics, sometimes within the network, and quite often across network interactions, which are prominent features of psychiatric disorders. So this has given rise to the third core principle that there are core prefrontal parietal control systems that can be dissociated into distinct brain networks or distinct roles in cognition. And I'll provide another illustration of this beyond just the dichotomy of task activated areas and areas that are relatively suppressed and challenging externally stimulus different cognition. And this is the dissociation of the frontal control system into two segregated systems, one called the salience network, and the other, the lateral frontal parietal central executive network. And so this study shows again, during a challenging cognitive task, the engagement of both these systems, but the de-engagement of the ventromedial prefrontal cortex. And now you can actually use tools like the ICA to decompose these coactivated systems into distinct components. And one of which is the salience network anchored in the anterior insula and the anterior cingulate cortex, which is very distinct from the dorsolateral prefrontal cortex and the posterior parietal cortex that's engaged during working memory. So these systems are, these areas are all active during complex cognitive tasks, but they are intrinsically quite distinct. And so this dissociation actually has allowed us to study what the differential roles of functional control systems is. And, and this body of work has led to the highlighting of three core networks, which are foundational to trying to understanding cognitive and affective dysfunction and dysfunction of cognitive control systems in psychiatric disorders. And I've already alluded to the default mode network, the salience network, which I just showed you is dissociated from the central executive network or the lateral frontal parietal network, FPN. And the thing to keep in mind is that we are actually trying to understand these systems and study them in the context of what they actually do during cognition. And a wide range of studies has shown that these areas that are highlighted as being part of the salience network and the frontal parietal central executive network are commonly activated in a wide range of cognitive control tasks, whether it be the stop signal task, the go-no-go task, the flanker task, and so on. So now there's a, there's a body of work that actually has led us to propose that there's a specific function of the salience network, which is really quite important in the context of psychiatric disorders. And the general idea is that this, the nodes of this network are activated across a range of cognitive tasks and affective tasks. And they're among the most extensively and commonly activated regions across all neuroimaging studies. And from the notion that these areas are activated by a range of not just cognitive tasks, but biologically salient stimuli, such as fearful stimuli or painful stimuli or disgusting stimuli, we asked the question, does a system that's designed to detect biologically salient stimuli also play a role in cognitive and affective control? And the answer is yes. And in a series of studies where we looked at causal signaling mechanisms using advanced computational procedures, we found that the frontal insular cortex is a causal hub in the sense that its activation causes these other task-related areas to either get activated or suppressed. And it's the largest causal signaling hub that they've been able to identify across virtually every cognitive task that we've looked at. And so here's just an illustration of that. It's across three different common cognitive control tasks. The different versions is the stop signal task, which involves response inhibition. You're given a go cue and there's a stop signal that asks you to withhold a response. The flanker task where there's congruous and incongruous stimuli, and another version of the stop signal task. And you can see, if you look at what is the flow of information in this system between the salience network, the frontal parietal network, and the default mode network, it's the anterior insular that sends out the largest signal across all of these tasks as illustrated here. So this has given rise to the notion that the salience network, sometimes referred to also as the singular-percular network, is critically involved in the ability to detect and attend to salient goal-relevant events in a flexible manner. And its impairment to this that I will illustrate later is a significant factor in cognitive control deficits in psychiatric disorders. So the salience network is a psychiatric disorder. And this is a system to reiterate anchored in the anterior insular and the anterior singular cortex. And it plays a crucial role in this process of responding to salient stimuli through rapid detection of goal-relevant events and facilitation of access to appropriate cognitive resources, as well as suppression of systems which are not actively required for the task and goal at hand. And so within this kind of framework of what the salience network does, and we'll get to the relevance for psychiatric disorders, it's a bottom-up saliency detector at a level that's really relevant to goal-directed behavior. This is really somewhat different from what's visually salient, for example. And that's a really different system in terms of, you know, detecting contrast and so on. This is not that kind of saliency. It's a saliency related to things that are personally irrelevant that you are obliged in a goal-directed sense to respond to. And the system plays a key role in switching between systems to engage the attentional and working memory system while down-regulating systems like the default mode network when you have to actually attend to something in the external environment. So you can imagine a scenario where dysfunction to the switching system, which is going to impair the ability to come out of your internal modes of thinking to respond to the external world. And this is kind of a core anchor for understanding psychiatric disorders in our framing. And then there's access to the motor system through the anterior cingulate cortex and response selection. And there's also an internal physiological reactivity to a saliency event having occurred by means of the interactions of the anterior and posterior insula. And the posterior insula are very strong links to the physiological and autonomic reactivity monitoring system in the hypothalamus and other subcortical areas. So at the end, this large body of work on these three networks in particular has emphasized the critical role of the salience network in attending to both sensory and it also turns out internal salient events, which then causes the switching on of cognitive resource for task performance. For example, if you have to keep information in mind and manipulate information during working memory, but it also then suppresses the DMN, which is more and primarily involved in self-referential mental activity. Now, there are new updates to the thinking about what the DMN does, but this is a canonical process and the DMN can swing back into action in terms of having detected and the ability to remember salient events. But primarily, this is the early process that actually occurs. So with this background in mind, and so we've developed a theoretical framing for thinking about how brain networks are organized. What are the key networks involved in cognitive control? And now the goal is to try and understand how dysfunction in these systems and their interactions can give rise to transdiagnostic features of psychiatric disorders, as well as individual phenotypic features that are manifest in specific psychiatric disorders ranging from autism to depression. So that's what I want to turn my attention to now in the second part of the talk. And I spent a fair bit of time trying to build a background and framing because there's a lot of work that has been going on in the field of brain networks with multiple different methodologies. And it's really important to have a theoretical anchoring because without it, the brain science literature is somewhat nebulous. And as study after study does different uses different types of methodologies and the stories can rapidly diverge unless we have a theoretical anchor. And so that's the reason that I've kind of spent a fair bit of time laying out this core architecture in the context of the triple network model and the role of different functional systems. There are other ways of thinking about it too, but I want to show you the really the value of this effort in terms of trying to understand aberrancies in cognitive control systems in psychopathology. And the general underlying principle, apart from the way the functional system is organized, of course, is that is the point I mentioned earlier is that almost all psychopathologies are dysfunctions of cognitive and affective control, the ability to bring cognitive control systems into to bear in a goal relevant manner. For example, to switch from a state of rumination to performing a task, executing a motor movement or processing a stimulus in a particular way. And so I want to go back to this general notion that I mentioned of each area having a different function, each brain area, because of its different functional connectional fingerprint. And this is kind of illustrated. Now, although I mentioned that there are no gross abnormalities in most psychiatric disorders, structural abnormalities, it is the case that most of the disorders that we study have some level of subtle psychoarchitectonic deficits. So the expression of different excitatory neurons, inhibitory neurons in schizophrenia, ER imbalance, and expression of one economal neurons in autism, the expression of different cell types. So all of these gives rise to the notion that there might be regional psychoarchitectonic deficits. And now, if you have such a deficit in one area and its connections are also impaired, you can see that it's going to impair a range of information processing. And moreover, if those deficits appear in core areas which are, whose engagement or disengagement is obligatory to task performance, the so-called functional hubs like the dorsolateral prefrontal cortex or the posterior cingulate cortex, you're going to get a range of cognitive and affective dysfunctions because you're not able to engage these core control systems. And that's kind of one of the principles underlying, you know, why we've kind of built this model for understanding cognitive control and how impairments to these core systems can result in a wide range of psychopathologies. So before I illustrate specific examples in individual disorders, I want to give you a sense of how this notion of the core cognitive control system has led to a proposal of a triple network model for major psychopathology. And at the core of this is impairments in the salience network, which is anchored in the anterior insula, the posterior insula, and the anterior cingulate cortex. And the idea is that weak mapping of salient stimuli, so a task-relevant event, an oddball event, or a stimulus that's asking you to withhold a response, does not gain access to the system because it's already over-engaged in some other process like rumination, for example, and depression. It's also sensitive to internal inputs. And so if there are impairments in those salient inputs, either because it's overactive or underactive in response to external stimuli, and potentially even overactive in response to internal stimuli, for example, an example would be a hallucinatory event, which is now seen as a salient event when it's not really manifest in the external world, can alter system dynamics in a fundamental way. An example of not just weak saliency, but exaggerated saliency would be a drug cue, for example. So the same cigarette or a drug would be just a visual stimulus for many of us, but it would take on a significant saliency for an individual with drug addiction. And so what happens is that that system, that stimulus then gains preferential access to the salience network, which I mentioned plays an important role in switching between other systems, in particular, these two core working memory and internal self-referential systems. And these impairments, either through the weak or exaggerated mapping of external stimuli, internal events, can result in a range of cognitive deficits because the system is not able to come out of states to engage in either process effects, relevant externally driven stimuli or internal mental events. So another example here would be in the case of major depression and rumination, which I'll show you examples of which result in exaggerated internal connectivity of the default mode network. And if this system is really hyperconnected, an external stimulus, our engagement of the salience network is not sufficient to ramp down the response of this default mode network. As a result, access to working memory resources is more limited. And this is the general idea. And this is a principle, a model that I've developed in the last decade or so, which has been finding a lot of traction across multiple disorders, both trans-diagnostically as well as in relation to individual clinical phenotypic features associated with each disorder. And so I want to illustrate some work from a range of meta-analyses, which are showing that the same cognitive control networks show up in a lot of psychiatric disorders. And here is results of a structural imaging study of individuals with generalized anxiety, depression, schizophrenia, and what really turned up across all these disorders, the common areas of dysfunction were the anterior insular cortex and the anterior cingulate cortex, the ventromedial prefrontal cortex, and the posterior cingulate cortex, which form core nodes of the salience and default mode networks. And this is illustrated here with respect to not just structural imaging, but also functional imaging, again, showing the impairments of these core systems occurs. And it's really remarkable that it's really these three systems that come up over and over again in a very trans-diagnostic manner. And here is another meta-analysis across thousands of individuals aggregated in the study that was published recently, again, showing altered connectivity or architecture of the default mode network, its connectivity with the frontal parietal network and its connectivity with the salience network. And the challenge now, now that we know that these core systems are dysfunctional is to understand how they become dysfunctional and how they become dysfunctional in each individual disorder beyond this notion of trans-diagnostic impairments. So what I want to do in the remaining part of the talk is give you examples from our work and those of others in specific psychiatric disorders. And I'm going to go through the developmental span to show you that these features appear across the lifespan, across different types of clinical manifestations, starting with autism spectrum disorder. And in this kind of area where we've done a lot of work in trying to understand early development and early development of psychiatric symptoms associated with social cognitive dysfunction, as well as repetitive behaviors is to try and understand circuit level dysfunction and then use this information to inform treatments. But the goal, the first goal is to really kind of play out in a and discover significant replicable effects in these systems. So this is a study that we published a few years back. So we identified functional networks. And the thing to note here is that these networks can be readily identified in children with autism. And so these are the similar canonical networks that I alluded to earlier, the salience, central executive, and posterior default mode network. And then we asked what systems are dysfunctional and how are they dysfunctional? And we found that some of these systems are actually hyper-connected, which was a little surprising in autism. But it's actually quite consistent with the model of EI imbalance, which suggests that there would be exaggerated local circuit and distal connectivity. So it's not really that surprising from a physiological point of view. And so then we asked which systems most clearly distinguish children with autism from controls, and here's our classification analysis. And it turned out that the salience network was a system that most clearly distinguished the children with autism from patrols. And then we asked what is the cognitive function or behavioral function that's associated with this dysfunction? And it turned out to be restricted and repetitive behaviors because this insistence on sameness, it engages the same system, it's hyper-connected, and it's difficult for the system to come out from it and respond to other salient events to which the child has to respond to. And in this recent study, which has just been accepted for publication, we've taken this a little bit further in terms of asking how these three networks interact and do their aberrancies to identify individual states and their aberrancies in interaction. And to ask whether, to dissociate different aspects of restricted repetitive behaviors. And in particular, the two different clinical symptomatic features of this phenotypic feature of autism in terms of insistence on sameness, which is more on a deficit on the cognitive side and repetitive motoric behaviors, which we hypothesize to be more related to motor circuit dynamics. And so what we did here was to identify time-varying connectivity, how the cross-network interaction takes place. And we identified significant deficits in cross-network interaction and the presence of different states. And so the way these systems interact between each other is highly dysfunctional. And we've shown this both in the cognitive control system that I alluded to earlier, but also a motor circuit, which is impaired, which is anchored in the primary motor cortex, the serradellum and the basal ganglia. And the important thing is that we're now able to actually dissociate the, do a double dissociation between the cognitive aspects of the repetitive behaviors, which is circumscribed interests and insistence on sameness from the motoric aspects, such that the cognitive control system deficits that I mentioned are related to the impairments in these two cognitive domains, whereas, but not the motoric domain, whereas the motor circuit impairments are related to the motoric deficits, but not the cognitive control deficits. So we've taken our initial work on identifying a dysfunctional system and then said, how does it relate to interaction of the salience network with other systems, in particular, the frontal parietal network and the default motor network on the cognitive side and motor circuits in relation to motor behavior. And we see that we can actually dissociate them. And this is really important for, important clinically because it tells you which system, that there are distinct systems and that you should target particular cognitive control systems for certain phenotypic features and other systems anchored in the cortex and basal ganglia for repetitive motor behaviors. And at some point they are related, but it's nice to see this dissociation. So what we've also done is not really understand, not just focus on intrinsic responses, but also if a circuit is hyper-connected, is it, does it fail to respond to external stimuli in particular here in this oddball task where you have to monitor for a deviant face and ask, does the network up or down regulate appropriately in relation to clinical symptoms? And what this shows is that higher repetitive behavioral clinical phenotypic weaknesses are associated with reduced ability to modulate the intrinsic circuits in response to task relevant goals. And this again illustrates the general principle that a hyper-connected system may not respond appropriately to external goal relevant stimuli. And this inability to modulate these networks is then a proximal cause and underlies a direct link to clinical phenotypic features. So this is just to kind of summarize that it illustrates that we can actually identify, use the same general principles to study these core networks that are impaired and their impairments in autism and the dysfunctional intrinsic dynamics leads to inflexibility of the system and the degree of inflexibility of the system is related to clinical phenotypic features. So I know I've presented materials related to autism but the general idea is the same. A hyper-connected system is inflexible. It does not respond to stimulus in a manner that's relevant to goal-directed information processing. And these features can predict clinical symptoms. So we've extended this work to ADHD as well and I'll provide you with some examples. The main takeaway here is that, again, it's a dysfunction of these core cognitive control systems that's impaired. And we've studied it intrinsically and the findings are really quite robust and the challenge is to try and understand it in the context of different types of cognitive and attentional tasks. So I won't go through all the details of these studies but I want to present you with the general idea and why it's important in the context of the models that I've laid out because it really, all roads lead through Rome and that's kind of the underlying theme here. So ADHD, of course, characterized by inattention and impulsivity, inattention, again, a cognitive control construct and so the general idea is that dysfunction in the cognitive control circuits should then be an important factor in inattentional symptoms of ADHD. And this is a series of studies. Again, these networks are easily identifiable. It's not that they don't exist but their internal structure, how they're internally connected within the network and how they interact with other systems, both intrinsically and during cognition is really what's at the core of these impairments. And so here in this study, we've looked at how the salience network interacts with the frontoparietal CEN and the DMNs. The general idea is that there should be increased coupling of the, this is the correlation between the time series that reflect each of these networks and the idea is that you have a network interaction index which is how strongly correlated the salience network in the frontoparietal network are and how strongly and inversely correlated the CEN and DMN are so which tells you, it gives you a measure of how well it's able to up-regulate and down-regulate systems in an intrinsic manner. And what we found is that these measures are significantly weaker. We're able to replicate this results to the overall connectivity measure is actually weaker in children with ADHD and also predicts inattention in both, in two independent cohorts from the ADHD 200 cohort. We've looked at the dynamics. We found that, and this is replicated in a second cohort, is that this, so we actually use a hidden Markov model or a state clustering algorithm to see how the connections are changing over time and we have, we can identify states. There are many more states in children with autism as more highly volatile systems is what we've referred this to this as. And the overall connectivity profile of, as indexed by NII is weaker in children with ADHD but it's more variable over time as I alluded to. And this is again replicated in the degree of this volatility is predicts clinical symptoms. And it does so in a very consistent way across both cohorts and these measures of inattention and impulsivity. And so again, these core cognitive control deficit systems are dysfunctional in ADHD and shows again the manifestation of cognitive control system dysfunction as a core in these three networks as a core aspect of ADHD. So we now kind of taking the model of cognitive control systems, what they do, we developed this model and we've shown that it works trans-diagnostically as illustrated through the meta-analyses. And these are all studies coming from different disorders and we're showing specific examples of intra and cross-network interaction deficits in childhood disorders. And this is just kind of a summary of various findings in the ADHD study, kind of alluded to them as more volatile, variable effects in ADHD. And because of this volatility, we're seeing effects related to clinical symptom features. And we have other studies where we've tied this to task during stop signal tasks, but that's for another day. So I wanna illustrate other aspects of work that have evolved in adult psychiatric disorders. I wanna focus on, again, very common themes and replicated findings in schizophrenia and depression. And part of this is an extension of this general notion of psychosis as misattribution of saliency, which Kapoor and others more recently have developed. And we've taken this model of aberrant saliency, which was primarily initially ascribed to the subcortical dopaminergic system to links with the cognitive control systems. And we've asked whether the dysregulated cross-network dynamics of these three core networks can contribute to, may underlie the positive symptoms of schizophrenia. Again, related to effects which are cognitive in nature. And so we've done a very similar thing here in this study that I illustrated earlier. We've taken these networks, we've asked how they interact with each other over time, and we've defined states where there might be some part of the circuits might be connected and some parts where others might be. And we've asked, how is this dysregulated dynamics across various states of connectivity related to the positive symptoms of schizophrenia? And is there an index at all? Again, we see that there are many more states so states are volatile. These are participants, this is time. You can see that compared to controls, patients with schizophrenia are moving through multiple states over time. And this is shown here as the controls having two states. Again, very similar to ADHD, more volatile states in individuals with schizophrenia. And again, we can look at the time varying connectivity. It's the variability is the overall connection is weaker in individuals with schizophrenia, but the variability is higher. And together, these time varying features of inter-network connections predict psychosis as shown in this canonical correlation analysis. And we're able to identify the specific PANS positive subscale measures, which are most strongly related to the overall expression of psychosis. And this is a study we published a couple of years back in biological psychiatry. And this is a replication cohort that we got through the BSNIP consortium. And again, very similar profile with a very strong expression of conceptual disorganization in relation to cognitive control deficits and dysfunction related to psychosis. So the other piece of work I just want to briefly mention is that this also has anchor in terms of trying to tie in the earlier work of Sridhar Kapoor, which was more subcortical documentary related to these cortical control systems. There's a very nice PET study showing that salience network connectivity is related to mesolimbic dopamine function. And we think there's a very interesting connection between dopaminergic expression and the ability to regulate connections in these networks and also cross network links. And so that there's a significant promise in terms of trying to tie in various models of saliency that have evolved in the past decade or so. And so this is the general kind of idea, again, relation to our model that involves this and also tying into this dysregulated dopamine hypothesis. There are stresses which impact the organization of these core systems. There's incorrect aberrant salience attribution, which can come from inappropriate processing of internal stimuli, such as hallucinatory events, which then gain control over the dynamics in ways that are maladaptive. And they give rise to a potentially large range of cognitive dysfunction. And so this is kind of the general theme that's involved and this is just basically summarizing our findings. In fact, we can actually, I didn't mention that we can actually take the connectivity measures and predict patients from controls with a very high level of accuracy. But more importantly, also predict clinical phenotypic features associated with psychosis. So there's definitely a signature in this network that dynamic model of core dysfunctional aspects of schizophrenia. Now, finally, I'll turn my attention to depression where there's also for a slightly different reason, the same circuits systems are impaired and that I'll take you through a few studies. This was the earliest study of its sort where we looked at individuals with major depression and looked at the cingulate cortex connectivity. And we found that that again was hyper-connected and it was with the thalamic region. And this gave us an initial and the degree of this connectivity was related to the duration of the current episode and suggesting that hyper-connected circuits are a core feature of recurring episodes, of the length of episodes. And this gave us the idea that there's a lot of potential signal here in this core aspect of this circuit dysfunction in depression. And now I'm just gonna summarize very briefly the evidence that has accumulated over the years through meta-analyses of mainly intrinsic connectivity studies, again, showing that there's dysfunction in the DMN, there's dysfunction in the frontal parietal network, and there's dysfunction in the ventral attention network, which is related to the salience network. And this is a meta-analysis published a few, four or five years back. This is just a summary illustrating which systems are hyper-connected. And it turns out that the posterior cingulate cortex and the ventromedial prefrontal cortex are highly hyper-connected in individuals with depression compared to controls. There's also some evidence of hyper-connection to the dorsolateral prefrontal cortex, but also under connectivity with other systems. You can see that dysregulation of this core system is again kind of appearing over and over again in multiple disorders, including depression. So each for a different reason, but the underlying neurobiological systems are really quite similar, although the dynamics is really what distinguishes them. And because we don't have data from individual sites across all these disorders, it's not been possible to put a full picture together except through somewhat indirectly through meta-analysis. And this is again a very recent work showing default mode network hyper-connectedness in individuals with rumination, again, falling on the core nodes of the default mode network. And so this summary slide from Hamilton and colleagues really summarizes the role of the DMN and also the interaction with the subgenual cingulate. There's a affect-laden behavioral withdrawal, which is hypothesized to then over-engage the core nodes of the default mode network, in particular in relation to assignment of value to internal stimuli. And there's also other evidence for dysfunction of the system in relation to the salience network. But the core aspect here, at least in the context of rumination, which is really a nice way to kind of frame this in the context of depression, is this exaggerated connectivity within the default mode network. And finally, I want to turn to a body of work, which is now a prominent aspect of treatment of depression, major depression, and the transmagnetic stimulation studies showing that the TMS functions by attenuating depression-related hyper-connectivity within the default mode network. So this is pre-treatment. Controls, you can see that across many nodes of the default mode network core nodes, you're seeing reduced hyper-connectivity. And moreover, consistent with the triple network model that I alluded to, you're seeing it also modulates the interactions between the DMN and the FPN. And so that's somewhat hyper-connected. Connection is also diminished after treatment. And these are busy slides, but I just want to mention that these findings are really consistent with the model that we've laid out in terms of core deficits in these systems and highlighting the specific aspect that is dysfunctional in individuals with major depression. And then the final kind of finding here in the study was that the subgenual cingulate connectivity with the DMN and the FPN predicts treatment response. And you can see here that of this connectivity in individuals with strong response versus the poor response. And so there's a lot of signature here in terms of how these systems are hyper-connected and how that hyper-connectivity within the network and across networks can be diminished with TMS treatment and also predicts clinical symptoms. And how you can use baseline information to predict clinical symptoms. So this brings me back to the model that we started out with, and I'll conclude with a concluding slide that summarizes this whole view. We've seen a number of different examples of individual psychiatric disorders where weak or exaggerated attribution of saliency from external events or internal mental events causes disruption of the switching on or off of task relevant systems. And this is a model for trying to understand cognitive and affective dysfunction trends diagnostically. And I've also illustrated the relation to individual phenotypic features in each disorder. So I want to conclude what I've illustrated here, I hope is that brain networks are useful for characterizing functional brain architecture. This is not just in terms of intrinsic organization, but also in relation to cognitive control tasks. We've seen that we've identified cognitive control networks, which play a role in flexible and adaptation and adaptive allocation of resources for task performance, the signaling on and off of systems that must be engaged or disengaged. And these form the core for trying to understand how brain networks function in psychiatric disorders. And I've illustrated cognitive control network models of psychopathology, in particular the role of the salience network in the context of a triple network model of psychopathology involving the salience frontal parietal and default mood networks, and how the volatility and the inflexibility of these systems is related to individual clinical phenotypic features of various disorders. And so it's somewhat surprising then that we can have a unified model where a number of these disorders impact very similar systems, which then give rise to ways of unifying both transdiagnostically and also looking at individual differences in a RDoC sense. And so the overall view then is that dynamic network models uncover clinically relevant features of cognitive dysfunction in psychopathology. And indeed psychopathology really at its core is dysfunctional cognitive control systems and access to those systems in a context specific manner. And these are the general principles I've tried to illustrate here with these examples. And thank you very much for your attention and interest. Thanks. ♪♪

brain networks

psychiatric disorders

cognitive control

systems neuroscience

triple network model

salience network

frontal parietal network

default mode network

autism spectrum disorder

attention deficit hyperactivity disorder

schizophrenia