Mark E. Pflieger, Ph.D. - US grants

The proposed project will develop an optimal methodology for non-invasive estimation and localization of real-time neural activity generated within multiple intracranial sources. This methodology will integrate EEG/ERP inverse methods with anatomical information from magnetic resonance imaging (MRI) to construct optimal head models and to constrain possible source geometries. The commercial aim is to develop modular software that supports a variety of source and head models for the analysis of surface- recorded electrophysiologic data. Immediate applications are envisioned in basic and clinical research ranging from functional neuroanatomy to cognitive psychophysiology. This user research will likely produce primary clinical applications in neurosurgery, neurology, psychiatry and neuropsychology. Phase I consists of a volume conductor study followed by an inverse operator study. Two new methods are derived for fast computation of volume conduction transfer coefficients using head geometry data: source- to-scalp geometric fitting, and transform to the unbounded volume. Transfer coefficient errors will be measured using the boundary element method as a standard. Applying different methods to simulated data constructed from actual MRI and EEG/ERP data, the inverse operator study will reveal the effects of measurement errors produced by coherently colored background EEG, and interactions with approximation errors produced by simplified head models.

Ongoing development of multimodal functional neuroimaging has been fueling productive new lines of research, in basic systems neuroscience, clinical neurology and neuropsychiatry. The dominant modality, functional Magnetic Resonance Imaging (fMRI), which is based on the measurement of activity-related hemodynamic changes, has not attained the millisecond resolution required for imaging of fast neuroelectric activity. Magneto- and electro-encephalography (M/EEG) can achieve this temporal resolution, although with uncertain and variable spatial resolution. Our objective is to estimate fast neuroelectric activity in brain regions of interest (ROIs) from M/EEG, using a threshold based on calibrated signal discriminability and spatial resolution characteristics. ROIs may be obtained from structural or functional MRI data. We describe a new method, REGional Activity Estimation (REGAE), that differs substantially from existing methods of M/EEG source analysis. REGAE optimizes and calibrates the theoretical tradeoff between ROI signal discriminability and spatial resolution. These calibration curves permit the user to fine-tune a ROI signal detector for a specified decision criterion. The aims of this work are: (a) to implement REGAE prototype software, (b) to systematically characterize factors that influence REGAE discriminability and resolution, including sensor configuration, signal-to-noise ratio, and ROI location, and (c) to integrate REGAE with existing commercial EMSE Suite software (http://www.sourcesignal.com/sw-desc.htm). PROPOSED COMMERCIAL APPLICATION: The software and methods that we propose are non-invasive, non- radiological and relatively low cost addition to existing EEG, MEG and MRI systems, and provide information that is not currently available from these systems independently. The resulting software will have direct application in clinical and cognitive neuroscience research. If clinical value is demonstrated, systems based on this methodology may find applications in the areas of psychiatry, neurology and psychology.

[unreadable] DESCRIPTION (provided by applicant): We propose to develop, validate, and commercialize innovative software tools for the study of causal dynamics of large-scale brain networks in humans. These tools may be used to provide neurologists with a reliable diagnostic for locating epilepsy seizure foci, and to provide neuroscientists with new analytic tools for assessing neural information transmission in the human brain. During Phase I, analysis methods will be developed and implemented to estimate the causal connectivity among selected brain regions. These methods are based on the computing the causal information between time series that represent brain activity in selected regions of interest, using the REGAE (regional brain activity estimation) algorithm. A nonparametric test of statistical significance will be developed and applied to the casual estimates. The results will be visualized using a causal diagram, indicating the statistically significant causal paths between brain regions, and their characteristic lags. The methods will be verified using simulated data, and evaluated with EEG data from epilepsy patients, as well as with evoked response data from normal subjects. Upon completion of development and testing, the new tools will be integrated into our existing EMSE Suite product for commercialization. [unreadable] [unreadable]

[unreadable] DESCRIPTION (provided by applicant): The work in this proposal will develop and make available to researchers new tools for studying time-varying changes in event-related human electrophysiological signals, such as those that might occur due to learning, habituation, facilitation, adaptation, or pharmacological intervention. Conventional average event- related electrical potentials and magnetic fields permit non-invasive exploration of underlying brain processes with millisecond resolution, but do not allow the quantitative analysis of interactions between events in a straightforward way. The work described in this proposal will implement experimental task- related system identification (Volterra) methods that will permit us to describe quantitatively both the linear responses to events (which are related closely to conventional average waveforms) and the non-linear dynamics of event interactions (which are not addressed by conventional averaging). Algorithms will be developed, embodied in computer software, and evaluated using simulated and experimental data. The resulting software will form part of our EMSE Suite commercial software package. A fuller understanding of these time varying changes will extend our knowledge in basic cognitive neuroscience, as well as its clinical applications, such as schizophrenia and other disorders of brain dynamics. [unreadable] [unreadable] [unreadable]

DESCRIPTION (provided by applicant): Blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) is the dominant noninvasive modality for studying human brain functional localization in basic and clinical neurosciences. Although the three-dimensional spatial resolution of scalp-recorded electroencephalography (EEG) is ambiguous, its temporal resolution is roughly three orders of magnitude better than fMRI. Consequently, a key issue in imaging neuroscience is how to integrate EEG with fMRI. The absence of a reliable computational bridge linking EEG to fMRI is a critical barrier to an integrated spatiotemporal experimental investigation of human brain function. We propose to develop a data-driven approach to integration-detection and estimation of regional BOLD- related EEG (rBRE) signals-which is constrained by a simple (though extensible) functional model of neuroelectric-hemodynamic coupling. An rBRE signal is a spatially and temporally filtered EEG signal which, when transformed via the functional model, demonstrates statistically significant coupling strength and regional specificity. Concurrent EEG-fMRI datasets are used to tune spatial and temporal filters which maximize EEG-BOLD coupling based on a particular form of conditional mutual information. After detection, an rBRE signal may be estimated at the temporal resolution of EEG. After successful completion of Phase I, we will have implemented rBRE signal detection algorithms in prototype software, verified their correct implementation using quasi-realistic simulations, and studied the effects of initialization errors, SNR, and region size. In particular, we will have shown that it is feasible to detect regional BOLD-related EEG signals reliably in human data. PUBLIC HEALTH RELEVANCE: How to integrate EEG with fMRI data is an important issue faced by many cognitive, behavioral, and social neuroscientists who are practitioners of both modalities. If successful, the research and development efforts described in this proposal will position SSI as a leading, innovative provider of software for integrated EEG-fMRI analysis. In addition to supporting concurrent EEG-fMRI capabilities, the developed software will be useful to neurophysiology labs which have access primarily to EEG apart from fMRI. Scientists working with us are interested in this software for the study of neurological, neurodevelopmental and psychiatric disorders, as well as basic brain research.

DESCRIPTION (provided by applicant): We address the clinically and scientifically important problem of how to integrate electroencephalography and magnetoencephalography (EMEG) with non-concurrently acquired functional magnetic resonance imaging (fMRI) data for the noninvasive study of large-scale, resting-state brain networks in temporal frequencies ranging from 0.08 Hz to more than 100 Hz. The unified analysis of low frequency hemodynamic signals with spatially ambiguous electromagnetic signals has been a critical barrier to progress in this field, and is addressed by novel methods in this application. These methods exploit the spatial structure of dynamic correlations found in resting fMRI to inform an EMEG inverse technique which, in turn, exploits the highly stochastic nature of ongoing brain activity. Cortical parcellations based on fMRI connectivity will be used to inform two complementary EMEG source estimation methods: A maximum entropy source covariance source estimator (MaxEntCov) provides a globally consistent solution, and a dual vector beamformer (DVB) provides local estimates for possibly- correlated activity in two source regions. MaxEntCov and DVB, which start from opposite perspectives, are combined to optimize model parameters in an iterative process that is designed to provide a convergent solution. After convergence, the estimators may be used to derive various EMEG-based functional connectivity measures. Our computational methods will be verified using quasi-realistic simulations. Research utility will be evaluated using data from an ongoing project to study the process of conversion to the psychosis. Finally, the prototype multimodal resting-state analysis tools will be integrated with commercial EMSE(R) Suite software. PUBLIC HEALTH RELEVANCE: Resting-state connectivity studies using low frequency fMRI have been fruitful in basic and translational neuroscience applications. Neurophysiological measures (EEG and MEG) reflect a much wider dynamic range of brain activity, but suffer from ambiguous localization in the brain. This problem is addressed by novel computational methods described in this application, which use MRI to facilitate a convergence of evidence from two complementary estimators of neurophysiological brain activity. Prototype computational tools will be evaluated in a psychiatric research setting. Successful completion of our aims will demonstrate that multimodal resting state network tools are feasible, useful, and ready for commercial development.