R&D Projects

ACCES

ICA

Phantoms

Stroke

Sleep

ACCES

Anatomically Constrained Conductivity Estimation System

Accurate EEG and MEG inverse solutions require accurate models of electric current flow in the head.

While most researchers use structural MRI to represent head geometry, few have addressed the need for accurate representation of tissue conductivity.

We have developed a method for measuring the average regional conductivity of the four major head tissues: brain, CSF, skull, and scalp. The method is based upon injecting non invasive levels of electric current into the scalp using pairs of electrodes in the Geodesic Sensor Net, and measuring the potentials at the remaining electrodes. The unknown conductivity parameters are then determined using a nonlinear inverse search algorithm.

Initial feasibility was demonstrated using a four-sphere model of the head. We found that all four regional tissue conductivities can be retrieved accurately despite the low conductivity and shunting effect of the skull. The method is now being extended to include realistic head geometry using boundary element and finite element methods.

Future extensions will include full tomographic reconstruction of the local skull conductivity, since the skull plays a dominant role in determining current paths through the head.

Rigorous testing of our conductivity measurement method is being obtained in software and hardware. First, we verify that our calculations obey certain identities with other problems, e.g., the Helmholtz reciprocity theorem. Second, we verify that our system accurately retrieves tissue conductivities in certain known cases. The first case is a cylindrical phantom with three layers including a simulated skull (Figure 1). The second case is the USC human skull phantom. (Figure 2). Finally, we are performing experiments in human subjects and building a database of conductivity parameters for different head sizes and ages.

ICA

top

Evaluating Independent Component Analysis

The Theory

The EEG is composed of electrical potentials arising from several sources. Each source (including separate neural clusters, blink artifact, or pulse artifact) projects a unique topography onto the scalp, called "scalp maps." These maps are mixed according to the principle of linear superposition.

Independent component analysis (ICA) attempts to reverse the superposition by separating the EEG into mutually independent scalp maps, or components. This is an extension to principal components analysis (PCA), which has had a place in EEG research for years.

The difference is that PCA separates input data only up to second-order statistics (e.g., co-variance or correlation), while ICA separates data using higher-order statistics.

If the input data are composed of sources with Gaussian probability distributions, then PCA will completely separate the data. While little is known about the actual distributions of brain activity, ICA practitioners claim that the sources comprising EEG are not of a Gaussian nature.

Applications to Dense-Array EEG

Recent literature shows a variety of exciting studies applying independent component analysis to EEG and MEG data. Applications reported include artifact isolation and removal, single-trial ERP analysis, and isolation of epileptic spikes, as well as a denoising preprocessor for source localization.

EGI is investigating many issues regarding ICA. The end goal is to provide our customers with the technical knowledge required to make consistent and effective use of ICA.

Our studies are focused on usability of artifact-cleansed data, single-trial ERP analysis, reliability of algorithm convergence, and required computing power.

Phantoms

Digital Phantom

Both EEG and MEG inverse solutions are based on forward models of how brain sources generate scalp fields.

Although this fact is known by most EEG researchers, the precise relationship between brain sources and scalp fields is unknown.

To support theories about forward models held by researchers and clinicians alike, we are developing a digitial phantom software package, called the Phantom Head Simulator, which will perform forward calculations for EEG and MEG.

The code base for this simulator is the same as that used for inverse solutions, so its users simultaneously gain knowledge of the latter by using the former. The present implementation is based upon a four-sphere model of the human head.

Future implementations will include realistic head geometry via boundary element and finite element head models.

Stroke

Stroke Monitoring with Dense-Array EEG

The recent use of thrombolytic drugs to effectively treat ischemic stroke marks a milestone in the campaign against this common neurological disease.

Although a significant advancement, this treatment is not suitable for all patients, such as those who are experiencing a hemorrhagic stroke or those receiving treatment more than three hours after stroke onset.

Hospital staff can use computed tomography (CT) scans to easily rule out hemorrhages. More difficult, however, is estimating the therapeutic window of time.

Late treatment still brings the hemorrhagic risk of thrombolytic treatment, but it faces an increasingly uncertain state of the neural tissue. A rapid, inexpensive, and noninvasive cerebral monitoring technique could help hospital staff to determine neural-tissue state as well as to assess the effectiveness of thrombolytic treatment.

In this Phase I project, we will use dense-array electroencephalographic (EEG) technology to collect near-continuous, stroke-related EEG so that we can evaluate the ongoing pathophysiological evolution of the ischemic tissue. Spectral and coherence analyses of the raw EEG data will enable us to characterize EEG changes in these records.

The results from this work will go toward the development of a rapid, inexpensive, noninvasive, accurate stroke-monitoring device based on dense-array EEG.

National Institute of Neurological Disorders and Stroke. Phan Luu, P.I.

Dense-Array EEG for Stroke Localization

Stroke can now be treated.

As new interventions are introduced for cerebral thrombosis and embolism, emergency stroke evaluation requires rapid assessment of ischemic brain damage. Studies with both animal models and human neurosurgical monitoring have shown that the EEG provides an immediate reflection of cortical ischemia.

In this Phase II project, we will implement the dense-array EEG protocol in emergency stroke evaluation at two clinical centers. We will develop advanced software modules for visualizing the EEG pathology in stroke. Finally, we will conduct statistical tests with the clinical trials data to quantify the improvement in localization with dense-array EEG.

Localization of the ischemic lesion is integral to detecting a stroke with the EEG, because focal slowing can be differentiated from diffuse slowing due to other cerebral pathology. Even with advanced methods, EEG localization of a completed infarction will be inferior to that from MRI. Furthermore, CT may continue to be required for detecting hemorrhagic stroke.

However, if we can improve localization with dense-array EEG methods, we can improve the emergency detection of stroke, and thus optimize the excellent temporal resolution of the EEG in characterizing the dynamic time course of the pathophysiology of ongoing cerebral ischemia.

National Institute of Neurological Disorders and Stroke. Don M. Tucker, P.I.

Sleep

High-Resolution EEG for Neonatal Sleep Monitoring

Electroencephalography is a safe, noninvasive technique for monitoring neonatal brain function.

Acute and long-term EEG monitoring of infants has clinical utility for evaluating the effects of hypoxic ischemic insults and detecting the presence of epileptic activity. The information available in neonatal EEG can be greatly enhanced by using an adequate number of electrodes to accurately map the spatial distribution of the EEG without spatial aliasing.

The Infant Geodesic Sensor Net will be designed based on experimental studies and theoretical simulations of the required sampling density for scalp potentials on the infant head. High-resolution EEG methods will be developed that incorporate the unique features of infant skull anatomy to estimate the potentials in the brain of the infant.

The Neonatal Sleep Monitoring system will be developed suitable for a variety of clinical uses including monitoring brain function in the Neonatal Intensive Care Unit (NICU).

National Institute of Mental Health. Don M. Tucker, P.I.