EMSE Source MRI includes all of the functionality of EMSE Data Editor + as well as the ability to perform various inverse solutions using individualized BEM, FEM and spherical head models (in individualized coordinates or MNI space). EMSE Source MRI includes all of the functionality needed to read and segment individual MRI, construct individual realistic BEM or FEM head models, and visualize results in the context of individual MRI slices or with respect to 3D head and brain surfaces extracted from the individual MRI for each participant.  

EMSE SOURCE MRI may be used for discrete or distributed dipole source analysis from EEG or MEG data. Three shell spherical models or realistic individualized or standardized head models may be used for field calculations. 3D source estimates may be 3D rendered and visualized on realistic individual or standardized head and brain surfaces or in slices derived from individual or standardized volumetric MRI datasets.

Functionality

Data Domains

• Source estimation may be applied to EEG or MEG data in the time domain (e.g., raw data or event-related average), in the frequency domain (e.g., power spectral density of ongoing or resting state data), or time-frequency domain (e.g., event-related evoked or induced oscillations).

Source Domains

• Continuous domain: One or more equivalent current dipoles with artibrary positions and orientations within the brain compartment.
• Tomographic lattice domain: 3-vectors of dipoles with x-, y-, and z-orientations distributed on a regular 3D grid that covers the brain compartment.
• Cortical surface domain: Dipoles distributed on a cortical wireframe with unconstrained or surface-normal constrained orientations.   Cortical surface wireframes may be obtained via built-in image processing functionality.

Image Processing Toolkit

• Basic image processing functions include thresholding, inversion, region growing, masking.
• Convolution filters include median, difference of Gaussians, Kirsch, Prewitt, Sobel, Marr-Hildreth.
• Segmentation functions include region growing, edge detection, outlining, erosion, dilation, and assimilation. Stop markers may be set manually.
• Histograms may be obtained for entire images or selected portions (regions of interest).
• Wizards are available for common tasks, including segmentation and mesh generation (see below), to automate parameter selection and sequencing for complex procedures. A scripting language is available for program automation of repeated tasks.

Surface representation/Mesh Generation

• Automatic mesh generation using a “shrink-wrapping” algorithm tiles a selected surface (e.g. the cortex) with a polygonal mesh. 3D rendering is available with VISUALIZER.
• Cortical inflation may be obtained from the cortex mesh with cortex curvature information explicitly retained for visualization.
• Boundary element model calculations may be performed on the surface mesh representations using SOURCE ESTIMATOR.
• Mesh Generation Wizard lets you construct a tiled surface or volume from any region creted by MR Viewer. You may then view this wireframe in the Visualizer module.
• Mesh Reduction allows you to retain the topology of the wireframe while lowering the mesh order (tile size).

Segmentation

• The Segmentation Wizard is a new feature that automates tissue classification within an MR volume. Some features are described below.
• Brain Extraction Tool (BET) is an automated algorithm for finding the brain within a a volumetric dataset.
• Hidden Markov random field algorithm automates cortical tissue classification into disjoint regions (e.g., CSF, gray matter, white matter).
• Inner/Outer Skull and Scalp Wizards complete the picture, enabling you to generate a complete finite element model for the MR volume.

Head Models

• Forward fields generated by current dipoles in the source domain are calculated at the sensors using a model of the head as a volume conductor.  Standard values for head tissue conductivities may be adjusted.
• 3-shell sphere.  Digitized sensor locations (e.g. electrodes digitized from LOCATOR) or manufacturer-specified locations may be used to determine a best-fit center of a 3-compartment (intracranial, skull, scalp) spherical volume conductor model.
• 1-shell sphere: May be used for MEG or ECoG source estimation.
• Boundary Element Method (BEM):  3-compartment head models for realistic-geometry boundary surfaces (scalp, outer skull, and inner skull) obtained from an average head MRI (included) or from individual MRIs.  The BEM head model assumes nested, homogeneous, isotropic compartments.
• Finite Element Method (FEM):  The most general realistic-geometry head model available in EMSE.*  Supports more than 3 tissue types which need not comprise nested compartments.  FEM meshes are obtained from an average head MRI (included) or from individual MRIs.  [*MEG not currently supported.  Anisotropic conductivity tensors available by special request.]

Volumetric Reconstruction

• Isometric reconstruction by linear interpolation for volumetric datasets.
• Image renormalization to correct for density changes from slice to slice.

Image Display

• Multiple windows for individual slice data with user selectable zoom and pan.
• Volumetric data may be displayed as 3 orthogonal slices. Optionally a mouse click will align the images on a selected voxel (“warp-to-point”).
• A comprehensive palette of Image editing tools is provided, including 3D cursor tool.

Multimodal Imaging

• Functional imaging data (e.g. fMRI, PET) may be superimposed on structural image data. Pseudocolor display of statistical values with user defined threshold.
• Source estimation data, including discrete dipoles and distributed source maps, may be superimposed on structural image data. Dipole position and orientation are represented, as well as sensor (EEG or MEG) locations. Pseudocolor display of source map values with user defined threshold.
• Seeded dipoles may be obtained from functional imaging data with mouse cursor selection. Dipole locations may be selected for further source analysis. A mouse click will align the images on a selected voxel (“warp-to-point”).

Coregistration

• Source estimation data may be registered with structural data by a least squares fit of electrode locations and/or 3D-digitized scalp points (e.g. from LOCATOR) to the MRI-derived head surface.
• Functional images may be aligned with structural images by a least squares fit of head surface points.

Inverse Methods

• Spatiotemporal dipole modeling. Fit one or more dipoles to the data. By interacting with the data, the operator may use up to 32 dipoles to create source models. The dipoles are subject to a variety of user-controlled constraints (including free or fixed locations and/or orientations, mirror pairs, and maximal power constraints).  Multiple source waveforms may be estimated using pseudoinverse techniques.  Confidence intervals may be estimated.
• Multi-start is an objective method for fitting multiple dipoles that minimizes user interaction with the data.  The spread of dipoles in point clouds indicates spatial extent and robustness of the source solutions.
• Dipole modeling in Frequency and Time-Frequency domains. Dipoles may be fit directly in the complex domain to Fourier transformed data, permitting the estimation of generators of oscillatory activity directly. Source Estimation using wavelets is a powerful tool for both time and frequency domains.
• Topographic mapping such as sphere-to-plane mapping, channel group selection (for dipole fitting), time-frequency display, Neuromag vector magnitude display, auto-update latency changes, context menus and a modeless dialog that lets you work while leaving the dialog open
• Seeded Dipoles. Dipole locations may be obtained from fMRI or PET hotspots and used for source estimation to effect constrained dipole time series analysis.
• Beamformers and other local estimators. Local estimators reveal spatially constrained source activity.
• Distributed source models. Source estimates are obtained from instantaneous data samples using maximum likelihood and Laplacian smoothing (LORETA) or covariance statistics (PRoMS) for a set of dipole current sources fixed to a lattice. Related options include minimum norm and MUSIC algorithm. The resulting analysis may be viewed as a tomographic display, either for a single time slice, or for a sequence of time slices (4D analysis)
• Cortical surface restriction*. Distributed source estimates may be restricted to a cortical surface derived from the individual MRI.

Display

• Dynamic linkage of graphical display windows permits selections in one window to update others automatically.
• Discrete dipole location, orientation, and confidence intervals are represented in diagrammatic display (3 orthogonal planes).
• Tomographic displays are available for distributed source estimates. 6D displays optionally include position and moment for each lattice point.
• 3D rendering and animation of cortical surface source estimates is available through VISUALIZER.

Input/Output

• User Interface. A conventional Microsoft Windows style graphical user interface is used throughout. Dynamic linking of graphical display windows permits selections in one window to update others automatically.
• Number of channels. Up to a total of 256 channels (EEG and MEG) are supported.

File I/O

• Input/Output functionality of Data Editor + is included.
• Image data input: ACR/NEMA/DICOM, Analyze (including SPM and Med-X), Raw byte, raw word, Analyze, GE XIMAGE, Picker, MINC, Stimulate (fMRI), with more added all the time
• Source estimation input: SOURCE ESTIMATOR, BESA, or ASCII (dipoles).
• Output. Binary volumetric image files (.vmi)

Other I/O

• Graphics may be transferred to clipboard for use by other graphics programs or for insertion into word processors.

Documentation

• Printed. User manual, including file formats.
• On-line. User manual and tutorial available as context sensitive hypertext