-
Notifications
You must be signed in to change notification settings - Fork 0
EEG Seminars
Our EEG seminar series showcases state-of-the-art EEG analysis methods. Every year, two in-person seminars are organized, each featuring (international) speakers on a common topic. In 2023-24 the topics will be machine learning in EEG research (19-Oct-23) and single-trial EEG analysis (31-May-24). We aim to foster an inspiring and interdisciplinary environment for PhD students, Post-Doc's and PI's to discuss the exciting possibilities EEG has to offer.
Registration is free, but required.
Please register here.
click here for more
By allowing data-driven discoveries, machine learning can identify unexpected associations between brain processes and cognitive functions. It can
integrate and analyze diverse data types such as EEG, behavior and model parameters.
Speaker 1: Prof. Diego Vidaurre (Aarhus University, DK)
Diego is an Associate Professor at the Department of Clinical Health in Aarhus University, and honorary member of the Department of Psychiatry in Oxford University. He uses computational models to get a better understanding of the organization of spontaneous brain activity, characterise how the brain processes concurrent stimuli, and describe individual differences in how these processes occur.
Speaker 2: Prof. Jelmer Borst (University of Groningen, NL)
Jelmer is an associate professor of Computational Cognitive Neuroscience in the Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence at the University of Groningen. His main research interest is the development and improvement of analysis methods that connect computational (cognitive) models to neuroimaging data.
Speaker 3: Dr. Ingmar de Vries (Radboud University, NL)
Ingmar is a Post-Doctoral researcher in the Predictive Brain Lab (PBL), led by Floris de Lange. He uses EEG and MEG to study predictive processing of naturalistic dynamic stimuli (e.g., videos of action sequences) in healthy human subjects.
Full program
09:00 Lecture 1: Diego Vidaurre
10:00 Coffee break
10:30 Lecture 2: Jelmer Borst
11:30 Coffee break
12:00 Lecture 3: Ingmar de Vries
13:00 Closing
Thursday Oct 19, 09h00 - 10h00
Beyond the hidden Markov model: applications for decoding and prediction of behaviour
Brain activity is vastly multidimensional and complex. Not only across subjects, but also across experimental repetitions or trials, brain pattern show great variability that traditional methods are often limited to accommodate. As a way to find structure within this complexity, thus enabling efforts to find meaningful associations to behaviour, the Hidden Markov Model (HMM) has emerged in the last years as a general family of models that can be used across data modalities. In this talk, I will introduce the HMM in its most general form, and discuss two problems/solutions: (i) how the HMM can help us to describe the unique information that individual trials have above and beyond crude averages; and (ii) how we can use the individual subject HMMs to predict individual traits such as brain age or intelligence.
Thursday Oct 19, 10h30 - 11h30
Understanding Cognition One Stage at a Time: Discovering Cognitive Events on Single Trials
Cognitive science has long investigated processing stages. Where the traditional methods of Donders and Sternberg were limited by behavioral measures, recent methods have used brain measures to track cognitive processing more directly. Here I will present one such a method, HMP, which is founded on two theories of EEG generation. Both the classical theory and the synchronized oscillation theory state that significant cognitive events cause multivariate voltage peaks that are added to the ongoing oscillations. The HMP method discovers such peaks, and thereby the onset of cognitive stages in EEG or MEG data using hidden semi-Markov models. It integrates the information present across all trials of all participants, to ultimately identify multivariate peaks on single trials. I will first show that this method can identify processing stages in simple decision and memory tasks, and that it can track which cognitive stages are affected by experimental conditions. Next, I will show that we can measure the onsets of cognitive stages at the single-trial level, allowing for many new ways to analyze data: from more precise ERP-like components, to within-stage connectivity, to linking continuous predictors to stage durations across trials.
Thursday Oct 19, 12h00 - 13h00
In dynamic environments (e.g., traffic or sports), our brain is faced with a continuous stream of changing sensory input. Adaptive behavior requires our brain to predict unfolding external dynamics. While theories assume such dynamic prediction, empirical evidence is limited to static snapshots and indirect consequences. We developed a dynamic extension to representational similarity analysis (dRSA) that uses temporally variable models to capture neural representations of unfolding events across hierarchical levels (from perceptual to conceptual). This approach quantifies the match between a stimulus model at a given time-point and the neural representation at different (i.e., earlier, or later) time-points. We applied dRSA to source-reconstructed MEG data and demonstrate neural representations of observed (predictable) actions to be mostly predictive. Predictions encompass several timescales matching hierarchical levels of processing, such that high-level conceptual features are predicted earlier, while low-level perceptual features are predicted closer to real-time. Additionally, we manipulated prior stimulus knowledge at two levels by either inversion (up-down) or temporal piecewise scrambling of action videos and observe a feature-specific reduction of predictive representations and increased unpredicted post-stimulus representations. To conclude, this new approach demonstrates the predictive nature of neural representations of naturalistic dynamic input and provides support for predictive coding theories of the brain.
Single trial EEG analysis allows researchers to capture rapid and transient neural events that may be smoothed out when averaging across trials. This enables a more precise understanding of the dynamics of brain processes and their temporal relationships.
Speaker 1: Dr. Eelke Spaak (Donders Institute, NL)
"Single-trial analyses as a step along the way from lab to life"
Eelke is an assistant professor and associate principal investigator at the Donders Institute, Radboud University Nijmegen. He is interested in how our conscious experience is co-determined by patterns of neural activity and connectivity and incoming sensory information. His research uses MEG, (intracranial) electrophysiology, and (“NeuroAI”) computational modelling to study the mechanisms underlying the integration between the two.
Speaker 2: Dr. Romy Frömer(University of Birmingham, UK)
“Using deconvolution to control for component overlap confounds in EEG correlates of decision-making”
Romy is an assistant professor at the University of Birmingham. She interested in how goals shape information processing prior to action, as well as when evaluating action outcomes e.g. to optimize learning. She is particularly interested in understanding how metacognition supports this flexibility in cognition. She uses EEG and computational modeling to answer these research questions
Speaker 3: Dr. Michael Nuñez (University of Amsterdam, NL)
"Capturing noise in single-trial ERP analysis using simple decompositions and neurocognitive modeling"
Michael is an assistant professor at the University of Amsterdam. He uses mathematical modeling to understand the brain and mind. His two main aims are 1) to further understanding of decision making and cognitive differences across individuals and 2) to develop modeling techniques that better test "neurocognitive" theories with real data. He evaluates electrical signals recorded from the brain while also evaluating human behavior during experiments.
Registration is free, but required.
Please register here.
The EEG seminar series is organized with the support of the Flemish government and of the doctoral schools of Social and Behavioural Sciences (SBS), Natural Sciences (NS), Life Sciences and Medicine (LSM) & Arts, Humanities and Law (AHL) of Ghent University. PhD students have the opportunity to receive credit with their doctoral schools for their participation (Transferable Skills - cluster "onderzoek & valorisatie").
For any questions, please e-mail the organizers Roos Doekemeijer (roos dot doekemeijer at ugent dot be), Sven Wientjes (sven dot wientjes at ugent dot be) or Raquel London (raquel dot london at ugent dot be)
click here for more
Spatial filtering methods (including e.g. Laplacian, ICA, PCA and generalized eigendecomposition) are widely applied in research and industry to reduce data-dimensionality and increase SNR. They can be used to denoise and to analyze data and offer interesting opportunities to increase the power of our analyses.
Speaker 1: Dr. Natalie Schaworonkow (Ernst Strüngmann Institute, Frankfurt, DE)
Natalie is a postdoctoral researcher at the Ernst Strüngmann Insitute in Frankfurt am Main in Germany. She works with David Poeppel to investigate dynamics of neural oscillations in EEG/MEG/ECoG/LFP data, and loves to post beautiful animations of the phenomena she studies on twitter.
Speaker 2: Dr. Simon Geirnaert (KU Leuven, BE)
Simon is a mathematical engineer with an expertise in signal processing and data mining. His research focuses on using the EEG to decode the locus of attention, and improve the performance of hearing aids. As a science communication enthusiast, he develops video abstracts, writes science blogs, and loves to deliver workshops and presentations.
Full program
13:00 Introduction
13:15 Lecture 1: Natalie Schaworonkow
14:15 Coffee break
14:45 Lecture 2: Simon Geirnaert
15:45 Coffee break
16:15 Informal panel discussion with the speakers for PhD students
17:00 Closing with drinks and snacks
Frequency tagging probes the EEG for the level of processing of a certain stimulus. It can be used in fundamental research and also has industrial applications such as in brain computer interfaces. Our speakers will offer technical and critical insights on frequency tagging and the phenomenon of neural entrainment from varied perspectives.
Speaker 1: Dr. Christian Keitel (University of Stirling, Stirling, SC)
Christian is interested in how our visual system deals with ongoing dynamic stimulation as we experience it in our everyday lives. He also undertakes frequent excursions into the areas of multi-sensory processing, intrinsic brain rhythms and, most recently, relationships between rhythmic brain activity and other physiological signals (pupil diameter, for example).
Speaker 2: Dr. Anne Keitel (University of Dundee, Dundee, SC)
Anne's lab at the School of Social Sciences at the University of Dundee investigates the neural underpinnings of speech, music and rhythm processing. A recurring theme is the interplay of the oscillatory organization of the brain and (quasi-) rhythmic naturalistic stimuli such as speech and music.
Speaker 3: Dr. Noor Seijdel (Max Planck Institute for Psycholinguistics, NL)
Noor uses computational models of visual processing (artificial neural networks, models of reading), neuroimaging and behavioral experiments. She is pioneering the use of Rapid Invisible Frequency Tagging to study the role of lower frequency oscillations in sensory processing and audiovisual integration.
Registration is free, but required.
Please register here.
The EEG seminar series is organized with the support of the Flemish government and of the doctoral schools of Social and Behavioural Sciences (SBS), Natural Sciences (NS), Life Sciences and Medicine (LSM) & Arts, Humanities and Law (AHL) of Ghent University. PhD students have the opportunity to receive credit with their doctoral schools for their participation (Transferable Skills - cluster "onderzoek & valorisatie").
For any questions, please e-mail the organizers Roos Doekemeijer (roos dot doekemeijer at ugent dot be) or Raquel London (raquel dot london at ugent dot be)