|
|
|||||||
|
|
The Alzheimer's Disease
Center at the University of Kansas diagnoses and provides clinical care
to patients with Alzheimer's disease. In addition, the Center supports
and engages in clinical and basic science research and promotes education
related to the disease. In 1992, Dr. Charles
DeCarli, an associate professor of neurology at the university and director
of the Alzheimer's Disease center, developed a patented, Fortran-based
algorithm for quantifying brain volumes from magnetic resonance images
(MRIs). After discovering Visual Numeric's PV-WAVE at a National Institute
of Health technology fair, DeCarli shifted his quantification algorithm
to PV-WAVE's Visual Data Analysis (VDA) platform. The PV-WAVE application
of DeCarli's algorithm became the cornerstone of a $2.7 million grant
from the National Heart, Lung and Blood Institute (NHLBI) on "Cerebral
Vascular Risk Factors and Brain Morphology in Twins." The research grant
focused on structural brain changes in healthy, aging patients and patients
with Alzheimer's and vascular diseases. "This project is
one of the first to document normal changes in brain regions. We discovered
that people don't lose brain tissue evenly," said DeCarli. "Our research
documented the difference in brain morphology between men and women, and
the effects of estrogen on the brain. As people age, men lose more brain
tissue from the frontal lobe; women lose more in the parietal lobe and
the hippocampus. These brain patterns are the same with Alzheimer's disease." PV-WAVE's Flexibility Supports Complex Visual Analysis and Customization DeCarli selected
PV-WAVE because the tool doesn't force him to apply visualization matrices
in any specific way. "With PV-WAVE's built in flexibility, I can utilize
general purpose functions and customize my own widgets. PV-WAVE's language
is so easy to use, I can put complicated images together with very little
effort," he explained. DeCarli applies statistical
analyses to his data for visualization purposes. "PV-WAVE allows me to
remove nonbrain tissue from an image. I reorient or correct the images
so they're homogeneous," he said. In addition, DeCarli
customizes PV-WAVE widgets to view mathematical arrays. Because MRIs add
spatial shading, or "noise", to images, DeCarli needed to apply a customized
mathematical array to sample images to identify and correct the shading.
He accomplished this by modifying the PV-WAVE Region of Interest (ROI)
tool. This change allowed him to clearly identify brain versus nonbrain
regions within MRIs (see slide sample). The slide sample
shown here includes various brain images that have been corrected using
a PV-WAVE-based algorithm. The images quantify regional brain matter volumes
and the volume of abnormal cerebral white matter. The images in the first
row display the method of removing nonbrain tissue from the original MRI.
The operator traces the dura mater, distinguishing brain and cerebral
spinal fluid spaces from the inner-table of the skull. The ROI is converted
to a bit mask as seen in the middle image. The masked image with the skull
removed is shown on the right. The second row is
an example of an MRI artifact correction that improves the accuracy of
the quantification. The left image is the original, the middle image is
the correction map and the right image is the corrected image. "The third row displays
the segmentation method used to identify brain and cerebral spinal fluid.
An intensity-frequency histogram is generated from the brain image. The
two intensity distributions are fitted automatically using a nonlinear
fitting algorithm. The segmentation threshold is defined as the minimal
probability between the two-modeled intensity distributions. The fourth row is
an example of abnormal white matter signal identification. In this case,
there is insufficient information to identify two separate intensity distributions.
The brain matter distribution is therefore modeled as log-normal, and
intensities above 3.5 standard deviations are classified as abnormal signals. "PV-WAVE has enabled
me to implement and modify algorithms and analyze complex data sets easily.
My current project involves over 400 MRI images, which is the largest
published quantification series in this research area to date. PV-WAVE
allows me, as a single user, to update a complicated system easily," said
DeCarli. "PV-WAVE's strength is the ability to easily customize visualization
tools and to create new functions for analyzing data. I can also produce
reports by writing data from PV-WAVE to a comma-delimited text file, which
I import into Microsoft Excel® for further manipulation
and charting." DeCarli has received
telephone calls from researchers from around the world who have asked
how he was able to produce these types of images. He now collaborates
with national and international Alzheimer's researchers who are also using
PV-WAVE on their projects.
DeCarli's technical
papers describing the above manipulation of brain images using his patented
algorithms applied through PV-WAVE have been published in over 12 journals,
including the Journal for Computer Assisted Tomography, the Journal
of Magnetic Resonance Imaging and Neurology magazine. | ||||||
