Session 1: Machine Learning
Monday, May 21st, 3:00 pm-4:00 pm
Presented by José del R. Millán and Julia Berezutskaya
BCI: A Mutual Learning Perspective (José del R. Millán)
Real-time signal processing and decoding of brain signals are certainly at the heart of a BCI. Yet, this does not suffice for subjects to operate a brain-controlled device. In the first part of my talk I will review some basic machine learning components that facilitate user learning as well. I will also discuss studies showing that BCI is more than just decoding. I will then illustrate recent results involving people with disabilities and the efficacy of mutual learning in the acquisition of BCI skills.
Bio: Dr. José del R. Millán joined the École Polytechnique Fédérale de Lausanne (EPFL) in 2009 to help establish the Center for Neuroprosthetics. He holds the Defitech Foundation Chair and directs the Brain-Machine Interface Laboratory. He received a PhD in computer science from the Technical University of Catalonia, Barcelona, in 1992. Previously, he was a research scientist at the Joint Research Centre of the European Commission in Ispra (Italy) and a senior researcher at the Idiap Research Institute in Martigny (Switzerland). He has also been a visiting scholar at the Universities of Berkeley and Stanford as well as at the International Computer Science Institute in Berkeley. Dr. Millán has made several seminal contributions to the field of brain-machine interfaces (BMI), especially based on electroencephalogram (EEG) signals. Most of his achievements revolve around the design of brain-controlled robots. He has received several recognitions for these seminal and pioneering achievements, notably the IEEE-SMC Nobert Wiener Award in 2011 and elevation to IEEE Fellow in 2017. During the last years Dr. Millán is prioritizing the translation of BMI to end-users suffering from motor disabilities. As an example of this endeavour, his team won the first Cybathlon BMI race in October 2016. In parallel, he is designing BMI technology to offer new interaction modalities for able-bodied people.
Fundamentals of machine learning applied to BCI research (Julia Berezutskaya)
State-of-the-art BCI relies on advanced data processing techniques necessary for transforming recorded brain activity into a series of meaningful signals. These signals are interpreted by a computer and used for communication of the user with the outside world. This complex process can be broken down into a number of steps, such as data collection, data preprocessing, feature selection and classification. In our session we are going to cover the basic machine learning approaches used during each of these steps. We will briefly overview the theoretical and practical aspects of widely used machine learning tools, such as regression models, support vector machines and artificial neural networks. We will combine high-level explanation of the topics with demonstration of real BCI examples utilizing these machine learning techniques.
Bio: I am a PhD student at the BCI lab lead by Nick Ramsey at the University Medical Center in Utrecht. I am working on decoding of language related information from electrocorticography neural recordings. I have two master’s degrees: one in Cognitive Neuroscience and another in Computational Linguistics. I have a keen interest in data science and machine learning with my current focus being on artificial neural networks, Gaussian process and Bayesian inference.
Session 2: Advanced EEG Analysis
Monday, May 21st, 4:00 pm-5:00 pm
Presented by Donatella Mattia and Dora Hermes
Basic facts about advanced EEG signal processing: from brain activity to connectivity (Donatella Mattia)
Brain activity can be recorded by means of EEG electrodes placed on the scalp. The scalp EEG reflects the synchronized activity of groups of neurons (neural-scale field potentials) across cortical areas and one of the fundamental problem in neurophysiology is the identification of the sources responsible of brain activity. Over the last decade, a variety of new methods have been applied to EEG signal processing that allow for an increase in EEG spatial resolution (inverse problem; source estimation). More recent interest has been focused on functional connectivity estimation which appears to be among the most informative features derived from high density EEG data. Overall, these new methods require new reliable research tools to allow for routine processing of EEG data. Here we will summarize principles behind methods, tools and fields of applications referring to as advantages and limits.
Donatella Mattia, MD, PhD, Clinical Neurophysiology, Neuroelectrical Imaging and BCI Lab Fondazione Santa Lucia, IRCCS, Rome
Bio: I am an MD, Neurologist, and a PhD in Neurophysiology (“Sapienza” University, Rome, Italy). I am habilitated as Associated Professor in Neurology. Past research position as research assistant at Montreal Neurological Institute, McGill University, Montreal QC (Canada) focused on basic neural electrophysiology in “in vitro” animal and human brain slice preparation. Currently I run the Neuroelectrical Imaging and Brain Computer Interface Lab., at the Fondazione Santa Lucia, IRCCS, Rome, (Italy). My research is focused on applying advanced biosignal (EEG) processing methods to investigate the human sensorimotor and cognitive function and the design and validation (clinical trials) of EEG-based BCI technology in the field of AT and neurorehabilitation (functional recovery/plasticity).
Basic facts on EEG spectral analysis in relation to models of underlying neural activity and hemodynamic activity (Dora Hermes)
BCI systems have used many different features extracted from the EEG recorded intracranially or on the scalp. EEG pools signals across very large populations of neurons, generating a complex signal and one way to look this signal is through the power spectrum. Recent models have been developed to relate spectral power changes to underlying neuronal population activity. Here, we will summarize some of these models to better understand the (i)EEG signal and its relation to fMRI.
Dora Hermes, PhD, Brain Center Rudolf Magnus, UMC Utrecht, Utrecht, The Netherlands, and Department of Psychology, Stanford University, Stanford, CA, USA
Bio: I am a postdoctoral fellow at the Brain Center Rudolf Magnus, UMC Utrecht in The Netherlands and a visiting postdoctoral scholar at Stanford University, Stanford. My research is focused on advancing our understanding of the mesoscale signals measured in the human brain and relating these signals to the underlying neuronal circuit dynamics in order to identify biomarkers for neurological diseases and develop neuroprosthetics. I did my PhD at the Brain Center Rudolf Magnus, UMC Utrecht, The Netherlands and did a post-doc at Stanford University and New York University.
Session 3: BCI Implant User Needs
Monday, May 21st, 5:00 pm-6:00 pm
Presented by Jennifer Collinger and Spencer Kellis
User Priorties for Implantable BCIs (Jennifer Collinger)
In order for BCIs to gain acceptance, it is important that we develop technology that meets user needs and maximizes the risk/benefit ratio for the user. In this session, I will discuss user priorities and design criteria for implantable BCIs as gathered from multiple survey studies of people with spinal cord injury and ALS. Specifically this includes priorities for functional recovery, discussions of desired functionality, risk considerations, and target performance metrics.
Jennifer Collinger, PhD, is an Assistant Professor at the University of Pittsburgh and a Research Biomedical Engineer at the VA Pittsburgh Healthcare System. Her research interests include neuroprosthetics, preservation and restoration of function, and neuromotor rehabilitation. She is part of an active human neuroprosthetics research program that aims to restore upper limb function through an intracortical BCI.
User Motivation and Outcomes for Implantable BCIs (Spencer Kellis)
BCIs have shown promise during initial clinical testing, but there are few commercial options available for paralyzed persons and no direct benefits from contributing to research. So, why do participants participate? I will examine motivations for involvement, looking specifically at risks and benefits, informed consent, and social impact. I will also look at current progress toward clinical translation, as a context for the participant experience.
Spencer Kellis, PhD, is a Member of the Professional Staff at Caltech, Director of Engineering for the T&C Chen BMI Center at Caltech, and Research Assistant Professor of Neurological Surgery Keck School of Medicine of USC. His research interests include control and sensory aspects of brain-machine interfaces (BMIs), machine-intelligent prosthetics, and neural signal processing. At Caltech, Dr. Kellis currently manages technical aspects of a clinical trial for intracortical bidirectional BMI.
Thank you to the 2018 BCI Meeting sponsors