Autumn 2020 – Neural Dynamics Forum

All Neural Dynamics Forum talks during Autumn 2020 will take place online through Zoom, details as below.

Meeting ID: 932 7630 6396
Password: 815933
https://zoom.us/s/93276306396

1PM December 18th – Tobias U. Hauser (UCL)

Do we need a developmental computational psychiatry?

Many psychiatric disorders arise during adolescence, a time when the
brain undergoes fundamental reorganisation. However, it is unclear
whether and how the emergence of mental health problems is linked to
aberrant neurocognitive development. In my talk, I will discuss why it
is critical to understand (aberrant) cognitive and brain development if
we want to better understand how mental health problems arise. I will
present findings showing how psychiatric traits are linked to adolescent
brain myelination, and illustrate why computational neuroscience
approaches could be important in understanding psychiatric disorders.

1PM December 11th – Aleks Domanski (University of Bristol)

Investigating population-level multisensory integration for predictive coding in the primary visual cortex

Why can you hear your friend more clearly in a noisy bar by watching their lips move as they speak?
During sensory processing under ambiguous conditions, integration of multisensory information may improve the extraction of input statistics and increase the accuracy of predictions about upcoming information. The computational principles underlying such a boost in predictive coding and their vulnerabilities to disruption, notably in Autistic individuals, remain poorly understood. Alongside primary sensory afferents, the recurrently connected network of primary visual cortex (V1) receives modulatory input from other sensory and frontal brain structures. Indeed, previous work demonstrates that visually tuned neurons in V1 also respond to tones and noise bursts. Recurrently connected ensembles of neurons conveying mixed combinations of audio-visual variables could, as demonstrated in higher-order brain circuits, provide a powerful computational substrate to facilitate the nonlinear classification of ambiguous sensory input. However, this is unexplored in the context of viewing more dynamically complex naturalistic scenes.
Here, I will examine a large population calcium imaging dataset (1000~2000 cells) from mouse V1 to study how past and current multisensory input statistics are integrated by the circuit during ambiguous natural movie viewing to improve the fidelity of predictive coding.

9AM December 4th – Jihwan Myung (Taipei Medical University)

Multiple circadian clocks that are not always synchronized

Circadian clocks are biological clocks that maintain near 24-h periodicity with high precision. These clocks synchronize and make a robust clock when coupled. An interesting but often ignored feature of these clocks is that they do not always synchronize completely—sometimes by design. A functionally relevant case of close-to-synchronization can be found in the central clock called the suprachiasmatic nucleus (SCN), where its subpopulations deviate from phase-locking as the day-length increases as if the degree of synchrony served as a mechanism of seasonal time encoding. We then discovered that robust circadian clocks exist outside the SCN and they are not phase-locked with the SCN clock. Since the datasets needed to make these observations have high fidelity in time, i.e., the variables are dynamic and not static, experimenters need to understand the theoretical background on oscillator systems when designing experiments and interpreting data. Conversely, theoreticians equally need to appreciate the complexity of the biological system and imperfections in experimental approaches. We discuss some cases of circadian phase coordination and close-to-synchronization behaviors in the molecular, cellular, and tissue levels, and how these can be studied by experimental and theoretical approaches.

9AM November 20th – Woo-Young Ahn (Seoul National University)

Deep digital phenotypes using computational modeling, machine learning, and mobile technology

Machine learning has the potential to facilitate the development of computational methods that improve the measurement of cognitive and mental functioning, and adaptive design optimization (ADO) is a promising machine-learning method that might lead to rapid, precise, and reliable markers of individual differences. In this talk, I will present a series of studies that utilized ADO in the area of decision-making and for the development of ADO-based digital phenotypes for addictive behaviors. Lastly, I will introduce an open-source Python package, ADOpy, which we developed to increase the accessibility of ADO to even researchers who have limited background in Bayesian statistics or cognitive modeling.

https://ccs-lab.github.io/team/young-ahn/

1PM November 13th – Naoki Masuda (University at Buffalo)

Recurrence analysis of dynamic functional brain networks of individuals with epilepsy

Functional brain networks have been suggested to vary over time. We propose a new method to characterize dynamics of functional brain networks using the so-called recurrence plots (RPs) and their quantification. RPs and recurrence quantification analysis (RQA) were originally proposed for single nonlinear time series and have been applied to a range of dynamical systems and empirical data including neural signals. Here we propose these methods for dynamic networks, where recurrence is defined at the level of the functional networks, i.e., a network recurs to a past network if the distance between the two networks is sufficiently small. For resting-state magnetoencephalographic dynamic functional networks, we found that functional networks tended to recur more rapidly in people with epilepsy than healthy controls. For stereo electroencephalography (sEEG) data, we found that dynamic functional networks involved in epileptic seizures emerged before seizure onset, and RQA allowed us to detect seizures. The proposed methods can also be used for trying to understand dynamic functional networks in brain function in health and other neurological disorders.

http://www.buffalo.edu/cas/math/people/faculty/naoki-masuda.html

1PM November 6th – Rick Adams (UCL)

Beyond E/I imbalance – clarifying the fundamental circuit dysfunction in schizophrenia using biophysical modelling of multiple imaging paradigms.

Subjects with a diagnosis of schizophrenia show consistent differences from controls in neuroimaging paradigms such as resting state (rsEEG and rsfMRI), mismatch negativity (MMN) and 40 Hz auditory steady state response (ASSR). The underlying circuit changes causing these group differences are unclear, however, and it is not known whether the same abnormalities could underlie group differences in all paradigms. Nevertheless, it is widely hypothesised that schizophrenia involves a loss of synaptic gain – e.g. due to NMDA receptor dysfunction – and disrupted ‘balance’ between excitatory and inhibitory transmission in cortical circuits. Here we analyse a neuroimaging dataset containing data from controls (n=107), subjects diagnosed with schizophrenia (Scz, n=108) and their first degree relatives (n=57) each undergoing rsEEG, rsfMRI, MMN and 40 Hz ASSR paradigms. We use a variety of dynamic causal modelling approaches to estimate synaptic gain and other circuit parameters in auditory and frontal areas. We find some striking commonalities across paradigms, not just in synaptic gain in the Scz group, but also in relationships with symptoms and cognitive function. The potential for development of a model-based biomarker of synaptic gain is discussed.

https://iris.ucl.ac.uk/iris/browse/profile?upi=RAADA06

1PM October 30th – Michael J. Frank (Brown University)

Striatal dopamine computations in learning about agency

The basal ganglia and dopaminergic systems are well studied for their roles in reinforcement learning and reward-based decision making. Much work focuses on “reward prediction error” (RPE) signals conveyed by dopamine and used for learning. Computational considerations suggest that such signals may be enriched beyond the classical global and scalar RPE computation, to support more structured learning in distinct sub-circuits (“vector RPEs”). Such signals allow an agent to assign credit to the level of action selection most likely responsible for the outcomes, and hence to enhance learning depending on the generative task statistics. I will present experimental data from mice showing spatiotemporal dynamics of dopamine terminal activity and release across the dorsal striatum in the form of traveling waves that support learning about agency.

http://ski.cog.brown.edu/

Leave a Reply Cancel reply