A Data-driven approach to measuring conscious and unconscious brain function William Bosl , William J. Bosl (Pediatrics, University of San Francisco, Boston, MA ) C19
Introduction: Although neuroimaging has opened up new avenues for experimental studies of the brain in various states of consciousness, these can still be difficult to perform and expensive, limiting applications. We propose that the ever-changing pattern of electrical fields produced by the functioning brain is by definition a complex dynamical system, implying that mathematical theorems for measuring system properties apply. New methods for analyzing EEG time series, based on concepts of phase space reconstruction, allow quantitative measures of brain function for consciousness studies. A new generation of wireless, easy-to-use, high quality EEG devices makes it possible to relatively easily measure large numbers of subjects in experimental paradigms for studies of consciousness. Methods: Many nonlinear measures, or 'dynamical invariants' of the (brain) system, may be computed from EEG time series using publicly available software. These include several variations on the concept of entropy, Lyapunov exponents, correlation dimension, Hurst exponent, as well as measures of synchrony. A more general approach based on recurrence plots in principle contains all of the dynamical information available in the time series. A critically important point is that the computed dynamical invariants contain information that characterizes the functional state of the brain even if we do not yet know how to interpret that information. The relationships between the brain's dynamical state at any time, as measured by EEG and the computed dynamical invariants, and observable macroscopic states of consciousness as determined by clinical or other evaluations, can be discovered using machine learning algorithms. This provides an empirical approach to mapping functional brain measurements and observed conscious states. Results: We present an examples from our research in neurology and psychiatry to illustrate this data driven paradigm. In a search for very biomarkers of epilepsy, EEGs from children with Rolandic epilepsy, a condition where seizures primarily occur at night during non-REM sleep, EEG-derived measures (dynamical invariants) were found that indicate the existence of this epilepsy syndrome even in awake children. In another project, patterns of brain electrical activity were found in infants as early as 3 months of age that predicted a later emergence of autism. In both cases, human experts examined the EEG traces and found them to appear normal. Discussion: We suggest that a similar paradigm of measuring EEGs during various states of consciousness is an effective empirical approach to studying brain states associated with the observable concept of consciousness. In this way, a reference database of EEG-derived states corresponding to various states of consciousness can be constructed and shared by the consciousness research community. We are preparing to set up this database for research, and welcome help from the consciousness community. Research Opportunities and Challenges: Research to apply nonlinear analysis to EEG measurements and map these to consciousness states is in the earliest stages and wide open for discovery. Establishing a community-wide database for storing EEG measurements plus simultaneous information about consciousness state, relevant medical, psychiatric, or neurological conditions would create a research resource that would accelerate empirical studies of functional brain - consciousness relationships.