"Stimulate: A GUI based fMRI Analysis Software Package" (Article from: NeuroImage, 3(3):S607, June 1996)
John. P. Strupp
Center for Magnetic Resonance Research
University of Minnesota Medical School
2021 Sixth Street S.E.
Minneapolis, MN 55455, USA
INTRODUCTION
Stimulate is a GUI (Graphical User Interface) based fMRI
(functional Magnetic Resonance Imaging) analysis software
package. It offers a comprehensive set of fMRI analysis
tools integrated into a single package that was designed
for both convenient and flexible data processing. The
user-friendly GUI allows the user to simply "point and
click" with the mouse to modify analysis or display
variables. Activation maps can be calculated from the
fMRI data and overlaid onto aMRI (anatomic MRI) image
displays.
PARAMETER MAP GENERATION
The primary processing step in fMRI is the generation of
activation maps to identify areas of brain activity. An
activation map is a subset of the more general class of
parameter maps. A parameter map is the resultant image of
some temporal data reduction. The intensity value at each
voxel of a parameter map is a parameterization of a feature
of its time course. There are currently three analysis
methods implemented for parameter map generation: 1)
Student t-Test, 2) Cross-Correlation, and 3) Temporal
Statistics.
Student t-Test
The student t-Test method for parameter map generation
allows the user to specify multiple t-Tests consisting of
multiple control and stimulation periods. There are two
parameter options for the map: 1) confidence level, and
2) percent intensity change. The confidence level option
simply gives the result of the specified student t-test.
The percent intensity change option is used for activation
maps. A voxel is considered active when it exceeds a
confidence threshold for all t-tests specified. The
percent intensity change is that of the average intensity
of the stimulation periods versus the control periods.
Cross-Correlation
The cross-correlation method for parameter map generation
uses a specified model for correlation against the
intensity time course of each voxel in the functional
image[1,2]. Stimulate allows two ways to specify this
model, one simply uses periods to specify a "box-car"
function and the other uses an ascii file of a time
course. The "box-car" model is specified by high and low
periods corresponding to the experimental paradigm of
control and stimulation periods, respectively. This model
is not constrained to being a pure "Box-Car," but allows
for the sides to have a finite slope. Three types of
parameter maps can be generated: 1) correlation value,
2) time shift or 3) percent intensity change. All three
parameter types are reported for the time shift where the
maximum correlation coefficient occurs.
Temporal Statistics
The temporal statistics method for parameter map generation
simply calculates either the mean, standard deviation,
minimum or maximum intensity value over a specified set of
periods. The standard deviation map is useful for
localization of vessels [3], while the other parameter
types are useful for determination of background threshold
values.
IMAGE DISPLAY
There are three greyscale display modes: 1) stimulation,
2) anatomic, and 3) mask. In each display mode, the user
selects the planar offset and number of images to display.
An interactive LUT (Look-Up Table) tool is provided for
dynamic contrast adjustments.
The stimulation mode displays the fMRI (e.g. BOLD or flow contrast) images used in parameter map generation. For the anatomic mode, an aMRI set of images, typically of higher spatial resolution and T1 weighted, can be displayed to facilitate landmark identification. The aMRI and fMRI images are automatically aligned if different resolutions were used during their acquisition. Images for the mask mode are generated through algorithms specified by image processing operations. There are over 20 operations available from the following categories: data, threshold, logical, arithmetic and morphologic.
Stimulate uses a psuedocolor overlay for superimposing parameter maps on top of the greyscale images. A color palette editor is provided to select or modify the parameter map overlays for both positive and negative activation. The overlay can be limited to those voxels whose activation value are within a floor (minimum) to ceiling (maximum) range. In addition, a minimum connected component count can be specified.
A cine tool enables the viewing of a mosaic of planar images in a movie loop. This is particularly effective in highlighting physiologic or subject motion artifacts. Individual mosaic movie frames can be saved in TIFF format for photographic quality hardcopy. The aMRI or fMRI image volume with embedded activation maps can be saved in an isotropic BOB (Brick of Bytes) format for import into a volume rendering software package.
TIME COURSE PLOTTING / TEMPORAL FILTERING
Plots can be displayed of the intensity time course for
those voxels identified by a ROI (Region of Interest).
ROIs can be drawn using either a point, rectangle, or
polygon mouse tool. A zoom feature enables the accurate
specification of ROI down to the distinguishable individual
voxel level. The user can choose to view the time courses
of each individual ROI voxel or as an ensemble average.
The plot is displayed in a popup window which provides a copy of the ROI and image, plus a statistical summary of both the time course ensemble and the activation. The contents of the entire popup window can be spooled to a printer as a postscript file. In order to correct for physiologic noise or baseline drift, a temporal filtering module enables for the algorithmic signal processing of the time courses. Algorithm components includes Fourier filtering (low, high and band pass), arithmetic, statistical and mathematical morphology operations.
TALAIRACH COORDINATES
Stimulate provides a method for the localization of
activation in Talairach space [4]. This is accomplished by
use of an overlay grid and mouse position display. The
user must first calibrate the coordinates on a set of
cardinal view (transaxial, sagittal, coronal) images by
identification of the anterior and posterior commissure,
mid-plane and brain bounding box. From these specified
points, a transformation is computed which converts the
laboratory frame coordinates to Talairach space. Anatomic
and activation maps can be spatially normalized to these
specified talairach coordinates for inter-subject
comparison.
DISTRIBUTION
A SunOS version of Stimulate is available via anonymous ftp
at ftp.cmrr.umn.edu/pub. License keys are available
via the web at www.cmrr.umn.edu.
Work supported by NIH Grant RR08079
References