Supercomputing center helping researchers animate 'digital
humans' in almost scary detail
Here's looking at you
Monday, February 26, 2001
By Byron Spice, Science Editor, Post-Gazette
When it comes to displaying the innards of the human body in all
their glory, Hollywood is vastly more successful than science.
In last year's "Hollow Man," for instance, we saw Kevin
Bacon's body disappear layer by layer, exposing his circulatory
system, his muscles and finally his skeleton. In the classic 1966
film, "Fantastic Voyage," a miniaturized submarine carried
Raquel Welch and the rest of us through the blood vessels of an
ailing physicist.
The real-life counterpart to these films, the National Library of
Medicine's Visible Human Project, can seem dowdy by comparison. The
computerized images of two cadavers are impressive in their detail,
but can only be viewed in static, two-dimensional cross sections.
That makes it hard to discern three-dimensional shapes and spatial
relationships.
Computer scientists, however, are finally developing the tools
necessary to wring more meaning out of the massive Visible Human
dataset, which fills more than 60 CD-ROMs.
Medical students will soon be able to navigate through multiple
planes of the Visible Man and Woman, following the curve of the
spine, or tracing the optic nerves as they slant upward and criss-cross
on their way to the brain's visual cortex. Improved versions of
the computerized cadavers might show individual capillaries and
nerves.
The ability to peel off layers of tissue, as in Hollow Man, "will
be on the desktop in less than 10 years," predicted Brian Athey,
a biophysicist and bioinformatics specialist at the University of
Michigan School of Medicine.
In surgical simulators similar to the flight simulators used by
pilots, surgeons will be able to practice procedures on virtualized
patients based on the Visible Humans. They will use specially designed
surgical tools that will make it feel as if they are cutting or
pushing aside real tissue as they maneuver their instruments inside
the virtual patients.
And perhaps most amazingly, the cadaver-based Visible Humans may
eventually be re-animated. Blood will again course through the arteries,
the heart will pump and the hair-like cilia of the inner ears will
flagellate in response to sound.
"We're really going one step beyond the Visual Human,"
said Donald Jenkins, a former director of anatomy at Georgetown
University now working for the National Library of Medicine. "We're
creating a digital human."
Frozen, sliced, photographed
Though no match for the current gross-out movie hit, "Hannibal,"
the making of the Visible Human had grisly aspects.
Two cadavers were selected -- a 39-year-old man executed by lethal
injection in Texas and a 59-year-old woman from Maryland who died
of a heart attack. Under the supervision of Victor Spitzer at the
University of Colorado Health Sciences Center, the cadavers were
thoroughly X-rayed and imaged with CT and MRI scans. They were then
encased in blue gelatin and frozen.
They were then sliced up -- not in the way that dead mobster Richie
Aprile was fed into a meat grinder last season on HBO's "The
Sopranos," but very carefully and in exceedingly thin cross-sections.
The male was carved into 1,871 1-millimeter slices, the female into
5,189 slices, each a third of a millimeter thick.
As each slice was shaved off, it was photographed and the image
converted into digital form that could be processed by computers.
The slices could then be digitally recombined, providing cross-sections
that could be viewed from the side and the front, as well as the
original transverse cross-sections.
Since the Visible Male was released in 1994 and the Visible Female
a year later, more than 1,400 companies in 41 countries have obtained
no-cost licenses for the database, using it as the basis for books,
CD-ROMs and such products as computer programs for planning surgery
and radiation therapy. Images are available over the Internet.
Despite that early success, enthusiasm waned because of the limits
on how the data could be viewed, said Michigan's Athey. "It
was kind of a hunk of meat out there," he added.
"Unfortunately, one of the things that happened is that people
are drowning in data," said Terry Yoo, a computer scientist
at the National Library of Medicine. Most users don't have access
to computers big enough or fast enough to handle the huge database,
so they must content themselves with studying small portions of
the body or making do with the limited information they can glean
from studying a series of two-dimensional cross-sections.
Improving anatomy lessons
That's beginning to change. Yoo, for instance, is leading a team
developing software tools for "segmentation and manipulation"
that would help computers distinguish a particular feature, such
as a bone, so users could study its shape, see it in relation to
other bones, and see it move.
Several groups, including those at Michigan, Colorado and Stanford
University, are developing new tools while also taking advantage
of the next-generation Internet, which would move data 1,000 times
faster than today's Internet.
Michigan's Athey is working with the Pittsburgh Supercomputing
Center to adapt the Visible Human for use by medical students studying
gross anatomy. Students would use the computer program to complement
what they learn from dissecting a cadaver.
Dissection, by its very nature, is a destructive process, so students
lose information about spatial relationships as their dissection
progresses, Athey noted. The Visible Human could improve that understanding,
while perhaps also reducing the number of cadavers required.
Students would need to follow features in the Visual Human that
lie in multiple planes, said Art Wetzel, an expert on biomedical
modeling at the supercomputing center. "They need to range
all over the body at any angle," he added. That's not been
possible previously.
The spinal column, for instance, is not in a straight line, so
a single two-dimensional cross-section viewed from the front could
never display more than a portion of the spinal cord. And looking
at the heart and spinal cord only from the side or from the front
could never convey, say, the heart's proximity to the spinal cord
-- a juxtaposition that becomes obvious if viewed in a cross-section
that angles through the chest.
The key to providing students this freedom to choose what they
see and whatever angle they want to see it from -- and to provide
it nearly instantaneously -- requires keeping as much of the database
in a computer's working memory as possible, Wetzel said. The workstation
Wetzel uses for this purpose at the supercomputing center can hold
about 4 gigabytes of data in its working memory, but even that's
just a tenth of the Visual Human database.
Wetzel and his supercomputing center colleagues are developing
networking techniques that would allow 40 students to use the data
in a central computer, a task that includes finding ways to compress
the data so it can be more quickly transmitted over next-generation
Internet connections. "Networks are getting faster," he
explained, "but they will always be a scarce resource."
Not only would such a system reduce the demand for cadavers, Athey
said, but it would address the wishes of today's student to be able
to study at home, or to study at 2 a.m. Eventually, the same sort
of learning tools could also be made available at elementary and
secondary schools.
Addressing a teacher shortage
Making the Visible Human relevant to anatomy classes also helps
address the growing shortage of anatomy professors. Biology students
years ago began forsaking anatomy and physiology -- fields with
few prospects for research -- for the hot fields of molecular and
cell biology.
And even trained anatomists began dropping out because they could
no longer obtain tenured faculty positions, said Jenkins, who still
teaches gross anatomy at the Uniformed Services University of the
Health Sciences in Bethesda, Md. "It's been a sad story,"
he added, because medical students still need to learn anatomic
facts. With senior anatomists retiring, cell biologists often are
drafted to teach in their stead.
The prospects for enhancing the Visible Human, however, suggest
students will soon have some exciting learning tools, said Jenkins,
who oversees the Michigan project.
"It will be like 'Fantastic Voyage,'" he said. "You'll
take trips into the body in very dynamic ways. You'll be able to
take a walk up the aorta."
Researchers already are discussing repeating the Visible Human
Project with an additional cadaver. This would allow them to correct
for some imperfections in the original data sets and increase the
level of detail by shaving the body into even thinner slices, each
a tenth of a millimeter thick.
That would enable users to see individual capillaries and nerves
not visible in the existing Visible Humans, and to better discern
the edges of organs. Or, rather than go to the expense of cryosectioning
an entire cadaver, it may be possible to perform this more detailed
sectioning for only particular areas of interest, such as the eyes
or the ear.
As researchers get more adept at displaying the Visual Human data,
they are increasingly using it as a canvas on which they can display
additional information. At Michigan, they are calculating the stresses
and strain on pelvic bones typical of walking or birthing; Athey
said these stress patterns can then be displayed on the Visual Human
bones much like a weather map.
Athey is working with hologram pioneer Emmett Leith to make a three-dimensional
representation of the Visual Human data. As with a similar Surgical
Workbench being developed at Stanford, the 3-D digital human could
be used by surgeons who want to practice their skills or attempt
new techniques.
At Colorado, researchers are animating the Visual Human's knee
joint and using the head and neck visuals to model the act of swallowing.
Other projects under way are modeling and re-animating other parts
as well, a mix-and-match approach familiar to anyone who has seen
"Frankenstein."
"We're making the hydraulic system come alive," Jenkins
said.
|