Hans Liljenström

A neural network model of attention-modulated neurodynamics

Visual attention appears to modulate cortical neurodynamics and synchronization through various cholinergic mechanisms. In order to study these mechanisms, we have developed a neural network model of visual cortex area V4, based on psychophysical, anatomical and physiological data. With this model, we want to link selective visual information processing to neural circuits within V4, bottom-up sensory input pathways, top-down attention input pathways, and to cholinergic modulation from the prefrontal lobe. We investigate cellular and network mechanisms underlying some recent analytical results from visual attention experimental data. Our model can reproduce the experimental findings that attention to a stimulus causes increased gamma-frequency synchronization in the superficial layers. Computer simulations and STA power analysis also demonstrate different effects of the different cholinergic attention modulation action mechanisms.

Stability and Instability in Autonomous Systems

Behaving systems, biological as well as artificial, need to respond quickly and accurately to changes in the environment. The response is dependent on stored memories, and novel situations should be learnt for the guidance of future behavior. A highly nonlinear system dynamics is required in order to cope with a complex and changing environment, and this dynamics should be regulated to match the demands of the current situation, and to predict future behavior. If any of these regulatory systems fail, the balance beween order and disorder can be shifted, resulting in an inappropriate and unpredictable behaviour. I discuss how such “mental disorders” might be related to the structure and dynamics of any autonomous cognitive system.

Decisions and Downward Causation in Neural Systems

For any complex system, consisting of several organizational levels, the problem of causation is profound. Usually, science considers upward causation as fundamental, paying less or no attention to any downward causation. This is also true for the nervous system, where cortical neurodynamics and higher mental functions are normally considered causally dependent on the nerve cell activity, or even the activity at the ion channel level. This study presents a computational approach to decision making (DM) and downward causation in cortical neural systems. We have developed models of paleo- and neocortical structures, in order to study their mesoscopic neurodynamics, as a link between the microscopic neuronal and macroscopic mental events and processes. We demonstrate how complex neurodynamics may play a role for the functions of cortical structures. While microscopic random noise may trigger meso- or macroscopic states, the nonlinear dynamics at these levels may also affect the activity at the microscopic level.

Commentary by Hans Liljenström

For any complex system, consisting of several organizational levels, the problem of causation is profound. Usually, science considers upward causation as fundamental, paying less or no attention to any downward causation. This is also true for the nervous system, where cortical neurodynamics and higher mental functions are normally considered causally dependent on the nerve cell activity, or even the activity at the ion channel level. This study presents both upward and downward causation in cortical neural systems, using computational methods with focus on cortical fluctuations. We have developed models of paleo- and neocortical structures, in order to study their mesoscopic neurodynamics, as a link between the microscopic neuronal and macroscopic mental events and processes. We demonstrate how both noise and chaos may play a role for the functions of cortical structures. While microscopic random noise may trigger meso- or macroscopic states, the nonlinear dynamics at these levels may also affect the activity at the microscopic level. We discuss some philosophical implications from these studies.

Neurodynamics of Decision-Making—A Computational Approach

Decision-making is a complex process that normally seems to involve several brain structures. In particular, amygdala, orbitofrontal cortex (OFC), and lateral prefrontal cortex (LPFC) seem to be essential in human decision-making, where both emotional and cognitive aspects are taken into consideration. In this paper, we present a stochastic population model representing the neural information processing of decision-making, from perception to behavioral activity. We model the population dynamics of the three neural structures significant in the decision-making process (amygdala, OFC, and LPFC), as well as their interaction. In our model, amygdala and OFC represent the neural correlates of secondary emotion, while the activity of OFC neural populations represents the outcome expectancy of alternatives, and the cognitive aspect of decision-making is controlled by LPFC. The results may have implications for how we make decisions for our individual actions, as well as for societal choices, where we take examples from transport and its impact on climate change.

Study on Single-Channel EEG Pattern Induced by Acupuncture

A new approach for recognizing single-channel EEG pattern induced by acupuncture on acupoint of Zusanli was presented. EEG was decomposed by ensemble empirical mode decomposition (EEMD) into a small number of intrinsic mode functions (IMFs) based on intrinsic local characteristic time scale. The power, standard deviation, and analytic amplitude of first two IMFs expressed the difference between normal EEG and EEG affected by acupuncture. The two EEG patterns could be classified by extreme learning machine (ELM) with faster and better classification performance compared with BP. Similar classification accuracy of different channels indicated the global EEG pattern affected by acupuncture.

A biologically based neural network model for decision making

We present a neural network model, describing an adaptive decision making process (DM), under varying internal and external contexts [1]. The model includes the three most crucial structures in both emotional and rational aspects of DM: amygdala, orbitofrontal cortex (OFC) and lateral prefrontal cortex (LPFC) [2-4]. Neural activities in these structures represent emotional attitude, expectancy value, and rules towards the options (Figure u200b(Figure1).1). The DM is modeled at a level of mesoscopic neurodynamics, using biologically inspired neural networks [5]. Rational and emotional associations are encoded with cell assembly oscillations in all three structures, determining the value of an option, V(opt), as a product of the number of activated cells and the mean frequency and amplitude of their oscillations (Figure u200b(Figure2).2). A decision is based on the competition among stored patterns, using cosine similarity of the frequency vectors f→. The option with highest value will win the competition in each system: n n n nFigure 1 n nInteractions of the main neural structures in the decision making process: Amygdala, orbitofrontal cortex (OFC) and lateral prefrontal cortex (LPFC). Inputs from sensory modalities and from rational and emotional feedback. n n n n n nFigure 2 n nActivity of a cell assembly representing one of possible options. Red and green traces show the activity of the inhibitory neurons and the blue one the activity of all excitatory neurons. n n n n n nV(opt)|sopt|.⟨fopt ,Aopt ⟩=|sopt|.⟨fopt ,Aopt ⟩,∀opt=1,...,n n n n nThe emotional system is goal directed on the predicted value of the options. Emotional memory of various options are formed and stored in the amygdala. A prediction signal is generated in OFC and an emotional response is measured as a level of satisfaction. n nLPFC is a pivotal structure in the rational analysis and is based on habitual DM. Declarative and procedural memories, as the main components of this system are activated by external stimuli. The bidirectional interactions between maintained rules in procedural memory, rational attitudes towards different options, and stored factual information in declarative memory, lead to the selection of an option as a rational decision. n nThe integration of emotional and rational activities results in a final decision. When an action is executed as a result of a decision its experienced value is compared with the expected one, and stored in memory. Depending on the sign and the magnitude of the prediction error (expected value - real value), the stored emotional and rational attitude might be updated. n nThe experience of our decisions/choices are learnt and may influence future decisions. For any particular input signal, the final decision could shift, depending on internal and external context. Large delayed rewards have a lower value, compared to small immediate rewards. This fact can be included with the help of a hyperbolic discounting function, which models the exponential reduction of rewards in terms of time.

Study on the EEG Rhythm in Meditation

Meditation affects the brain rhythm significantly. Compared with non-meditators, the power of delta band was lower while high frequency band was higher for a meditator. Alpha band over the scalp was much more active in normal state for meditator with decreased dominant alpha frequency. Obvious transient process between normal eyes-closed rest and meditation was observed after EEG analysis. The active time and the power of beta and gamma band increased significantly in meditation. The inter-hemispheric and intra-hemispheric coherence beta and low gamma bands for meditative state were higher than normal state.

Phase Transitions in Mesoscopic Brain Dynamics – Implications for Cognition and Consciousness

Mesoscopic brain dynamics, typically studied with electro- and magnetoencephalography (EEG and MEG) display a rich complexity of oscillatory and chaotic-like states, including many different frequencies, amplitudes and phases. Presumably, these different dynamical states correspond to different mental states and functions, and to study transitions between such states could give us valuable insight in brain-mind relations that should also be of clinical interest. We use computational methods to address these problems, with an objective to find relations between structure, dynamics and function. In particular, we have developed models of paleo- and neocortical structures, in order to study their mesoscopic neurodynamics, as a link between the microscopic neuronal and macroscopic mental events and processes. In this presentation, I will describe different types of models, where the emphasis is on network connectivity and structure, but also including molecular and cellular properties at varying detail, depending on the particular problem and experimental data available. We use these models to study how phase transitions can be induced in the mesoscopic neurodynamics of cortical networks by internal (natural) and external (artificial) factors. We relate and discuss the models and simulation results to macroscopic phenomena, such as arousal, attention, anaesthesia, learning, and mental disorders.

Attention Modulation of Sensory Systems

Attention and arousal seems to enhance the efficiency of neural information processing. We use computer simulations of various sensory systems in order to investigate how the neurodynamics of these systems can be modulated for optimal performance in an unknown and changing environment. Using an inter-scale model of the insect antennal lobe, we demonstrate how attention modulation can change the sensitivity for sex pheromones. We also study how neural oscillations in mammalian olfactory cortex can serve to amplify weak signals and sustain an input pattern for more accurate information processing, and how chaotic-like behaviour could increase the sensitivity in initial, exploratory states. Finally, we use a neural network model of visual cortex area V4, in order to investigate potential cellular and network mechanisms for visual attention, reproducing experimental findings of attention induced gamma-frequency synchronization.