Guest talk: Ivan Zubarev "Convolutional Neural Networks for decoding electromagnetic brain activity"
Convolutional Neural Networks for decoding electromagnetic brain activity
Dr. Ivan Zubarev
Department of Neuroscience and Biomedical Engineering, Aalto University
Date: 1 November at 10:00-11:00
Functional brain imaging technologies allow studying the human brain activity at rest or in response to external stimuli non-invasively. Traditional analysis techniques rely on explicit biophysical modeling of thedata-generating process to identify components in the measured signal are associated with experimental manipulations. As the complexity of these measurements increases, data-driven machine-learningapproaches gain popularity allowing to probe whether brain states can be predicted from the data directly. These approaches hold great promise in for example identifying biomarkers of neurological and psychiatric conditions and developing brain-computer interfaces.
In this lecture I will describe specificity of brain imaging as an application domain of (deep) machine-learning techniques, introduce CNN-based methods for predicting brain states from the magnetoencephalographic (MEG) measurements, and briefly discuss the role of prior knowledge, and complexity – interpretability dilemma in deep learning.
Ivan has got his MSc in neuroscience from St-Peterburg University in 2013 and defended his PhD in technology (topic: “Developing machine-learning methods for the analysis of electromagnetic brain activity”) at the Department of Neuroscience and Biomedical Engineering (Aalto University) in early 2021. At the moment he is a postdoc at neuroimaging methods group at Aalto and a co-founder at Neubit Oy.