A computer model
of a functioning human could be the product of one of the largest
multidisciplinary projects ever undertaken. |
|
A view of the thorax of the human body based on Visible Human images. Drawing by Allison Baldwin (using thorax model by Eduardo Difilippo, Mark Lambert, and Ross Toedte). |
While digging holes in his vineyard one weekend
in October 1996, Clay Easterly conceived the seed of a grand vision
that is spreading like a slowly growing grapevine. Easterly, a health
physicist in ORNL's Life Sciences Division (LSD), saw the need for the
mother of all computer modelsa huge simulation of the structure and
function of the human body that would integrate smaller models of individual
organs, body processes, cells, and even neurons in the brain. The flesh
and blood of these models would be floods of data, ranging from digitized
anatomical images from the National Library of Medicine's Visible Human
Project, to known electrical and mechanical properties of human tissue,
to information on gene structure and function emerging from the federal
government's Human Genome Project.
"I saw it as a marriage of the new
biology and high-performance computing," Easterly says. "It's also a
way to bring many disciplines together to organize the flood of new
biomedical information, to solve complex medical problems, to identify
the medical research and data collection and interpretation tools that
are needed, and to drive the development of new technologies."
Easterly
had returned from a meeting at Quantico, Virginia, with officials of
the U.S. Marine Corps who are involved in its Joint Non-lethal Weapons
Directorate. The Marines are interested in developing non-lethal weapons
that are less harmful than the rubber bullets used now for such purposes
as keeping people from blocking food-bearing convoys on humane missions
to relieve starvation. One question raised by a Marine official at that
meeting was whether researchers could computationally simulate the response
of the body to rubber bullets and other kinetic weapons that may be
developed. Modeling human response was seen as being preferable to actually
testing it to determine if the weapons are non-lethal. "After our meeting
with the Marines, my Oak Ridge colleagues Glenn Allgood, Mike Maston,
Blake Van Hoy, and I had a crab and shrimp dinner," Easterly says. "We
talked around the idea of modeling, but it did not dawn on me then what
I would eventually consider. I think that the others may have thought
about something with the magnitude of the Virtual Human Project then,
but it just didn’t sink into me until the next day."
When he returned
to work Monday at ORNL, Easterly broached his computer model idea to
Barry Berven, associate director of LSD. Berven suggested that Easterly
set up a meeting with researchers who might want to pursue the vision.
Among the attendees were Allgood of the Instrumentation and Controls
(I&C) Division, Maston of the National Security Project Office. Richard
Ward of the Computational Physics and Engineering Division (CPED), and
Nancy Munro of LSD.
Ward and Munro revealed that they had both
been thinking along the same lines. In the spring of 1996, theytogether
with Keith Eckerman of LSD and John Munro of the I&C Divisionhad proposed
developing a model dubbed Physiological Human, using anatomic data from
the Visible Human Project and physiological models from ORNL's Dosimetry
Research Group and other sources.
|
Clay Easterly (sitting
at the computer) discusses the Virtual Human project of developing
computer models to simulate the body’s functions with Nancy Munro
and Richard Ward.
|
In an earlier project Ward, Brian Worley of
the Computer Sciences and Mathematics Division (CSMD), and Eckermann
had used Visible Human data to build a phantom (computational anatomic
model) for radiation treatment planning. The model, which used Monte
Carlo calculations, was developed to help doctors decide how to deliver
the maximum dose of radiation to a tumor while minimizing the exposure
of normal tissue.
The Visible Human Project is a 15-gigabyte
collection of digital images that shows the human anatomy in colorful
detail in a three-dimensional (3D) computer model. Available on the
World Wide Web since 1994, the collection consists of computerized axial
tomography (CAT) and magnetic resonance imaging (MRI) scans and photographs
of thousands of wafer-thin slices of two bodies donated to science.
The Visible Man was an executed Waco, Texas, convict who had killed
a 75-year-old man who surprised him while he was stealing a microwave
oven. The Visible Woman was a 59-year-old Maryland resident who died
in 1993 of a heart blockage that had no effect on the body’s appearance.
As Easterly and the other meeting participants
talked, the term "Virtual Human" emerged. They envisioned
a project that went beyond computer visualizations of body structure.
Their concept was to combine models and data to build a comprehensive
computational capability for simulating the function as well as the
structure of the human body. This capability will allow trauma simulations
and other applications.
In 1997 and 1998 Easterly, Ward, and Nancy
Munro advocated the Virtual Human idea among their ORNL colleagues.
Johnnie Cannon, director of ORNL's Office of Planning and Special Projects,
provided program development support and was one of the first to appreciate
the value of this vision. Easterly and Maston obtained some funding
from the Marines for preparatory steps leading to non-lethal weapons
research involving computer modeling. In 1998 Ward obtained some internal
funding for coding models of the cardiovascular system and developing
an easy-to-use interface. Ward and Ross Toedte of CSMD linked the model
to a Virtual Reality Model Language (VRML) graphics file of the human
thorax (the body region between the neck and the diaphragm containing
the heart and lungs). The user can click on and rotate the thorax on
the screen, viewing it in three dimensions.
Growing Support
In 1999 the Virtual Human Project took a major
leap forward. Easterly gained support for the idea from upper levels
of ORNL management. As a result, Ward and his colleagues received internal
funding for Virtual Human Project studies from the Laboratory Directed
Research and Development (LDRD) Program. Most important, ORNL Director
Al Trivelpiece embraced the idea. He told Easterly that the best way
to get congressional support for such a massive research program, which
could be on the scale of the Human Genome Project, was to get a favorable
recommendation from the National Academy of Sciences. But first the
Academy would have to agree to conduct a study.
On October 28, 1999,
at the instigation of Trivelpiece, a Virtual Human Workshop was held
at the National Academy of Sciences in Washington. The meeting was chaired
by Charles DeLisi of Boston University, who helped start the Human Genome
Project (and is on the LSD advisory committee). Attending the meeting
were scientific leaders representing the Department of Energy, Department
of Defense, National Institutes of Health, National Academy of Sciences,
National Science Foundation (NSF), National Aeronautics and Space Administration,
National Space Biomedical Research Institute, the White House’s Office
of Science and Technology Policy, several universities, four DOE national
laboratories, and the private sector.
"These representatives gave solid
support to the idea," Easterly says. "The meeting resulted in a recommendation
that the National Academy of Sciences perform a study to define and
scope the Virtual Human Project and make a report to Congress."
On November
8 and 9, 1999, the First Virtual Human Roadmapping Workshop was held
in Rockville, Maryland. It was organized by ORNL and co-sponsored by
the Joint Non-lethal Weapons Directorate.
The consensus of the two workshops
is that the Virtual Human concept is "an idea whose time has come,"
Easterly says. "This project is seen as necessary to help manage the
information explosion transforming biology and medicine by giving it
organization and infrastructure. It is beginning at the right time because
of the explosive emergence of information technology and supercomputing
capabilities and the ability to obtain more data from the body using
implantable sensors. But the project is in need of a very thorough and
systematic planning process."
So far, the leadership of the project
is coming from Oak Ridge. "We are beginning to think of the Virtual
Human Project as a resource," Easterly says, "because it will require
enormous databases, system integration, measurement systems, Web access,
computer hardware, computer infrastructure, model development, analysis
of individual biochemical measurements, and simultaneous analyses of
hundreds to thousands of measured variables. We expect thousands of
collaborators to make this massive multidisciplinary project work. It’s
an opportunity for biomedical researchers to put their results in the
context of the human body."
"We will need researchers representing many
different disciplinesanatomy, physiology, biomedical engineering,
electrical engineering, computer science, physics, biophysics, systems
engineering, information technology, and medical technologies," Ward
says. "We are teaming with experts in universities, medical colleges,
and industry to build the Virtual Human."
What ORNL Can Do
"ORNL
is a leader in developing the Virtual Human resource by integrating
the computer models," Easterly says. "We also will be involved in developing
specific models, such as a model of the respiratory system. And we will
develop implantable sensors, analytical instruments, and other methods
for improving data acquisition and interpretation."
For example, ORNL
researchers envision that the Virtual Human resource will give physicians
access to state-of-the-art, nonlinear analysis capabilities for rapid
remote analysis of patient electroencephalogram (EEG), electrocardiogram
(EKG), breath sound, and other types of data. A demonstration of this
capability is being planned by Lee Hively of ORNL's Engineering Technology
Division (ETD) using his patented algorithms for predicting seizures
from EEG data showing the brain's electrical activity.
Ward and his
colleagues are involved in both model development and model integration
projects. The goal of an LDRD project is to develop a Virtual Human
Integrated Respiratory System Modela computer simulation of the lung
and associated structures involved in generating and transmitting lung
sounds. Project participants include Allgood (I&C); Toedte (CSMD); Hively
(ETD); Nancy Munro (LSD); and Kara Kruse (CPED), who is a biomedical
engineer. The ORNL group is collaborating with Vanderbilt University
to incorporate a lung fluid transport model.
"We plan to build a model
that predicts lung disorders, such as emphysema and asthma, based on
changes in breath sounds," Ward says. "We will use patient data obtained
by Glenn Allgood at the Walter Reed Army Institute of Research using
the lung sound monitor he helped develop at ORNL. In a collaboration
with scientists at North Carolina State University and the Medical
College of Ohio, we are using computational fluid dynamic methods to
model the generation of lung sounds."
Here's a practical example of
how the ability to monitor lung sounds to detect disorders will prove
useful: The U.S. Army is interested in predicting whether a soldier's
lung has collapsed as a result of a puncture wound. If the chest wound
is large, a medic might not have access to the thorax area, so the best
way to get a diagnosis may be to place a sensor on the throat to gather
data on breathing patterns. The data could be wirelessly relayed to
the medic's wearable computer, which would then feed the data into a
model of the respiratory system for diagnosis of the lung injury, if
any. ORNL researchers and collaborators are building such a model, which
could save lives by telling a medic quickly what sort of trauma care
should be administered.
The lung disorder prediction model will be linked
to the Human Respiratory System Model of the International Commission
for Radiological Protection for predicting how much inhaled material,
such as beryllium or radon-daughter products, is deposited in various
parts of the lung. Visible Human data obtained from the National Library
of Medicine and data from human lung casts are being used for this project.
Ward is also leading an effort on model integration,
to help develop the computational infrastructure for the Virtual Human
Project. A major problem impeding the progress of this effort is that
computer models of various organs will be too big and complex to be
stored and executed on a desktop personal computer. The idea, therefore,
is to use computer technologies to allow the desktop computer user to
connect with large computer serversincluding high-performance
computersthat can run large models and feed the results in visual
form back to the user.
"We are
developing a Virtual Human Portal," Ward says. "This is a Web site that
can be customized to allow access by a desktop computer user to problem-solving
and modeling applications on a remote server."
|
The Virtual Human Java interface, which includes a 3D model of the human anatomy (not shown here), also displays a circuit for the left side of the heart and systemic vascular system (V.C. Rideout, 1991) and plots showing changes in left ventricle and capillary entrance pressures over time. This interface allows the user to run the model on a remote server using a Web browser on the user’s desktop machine.
|
For this purpose CPED's Dennis Strickler modified
an interface developed by Johnny Tolliver and Strickler for the Graphite
Reactor Severe Accident Code (GRSAC). The original code was written
by Syd Ball and Delphy Nypaver, both of the I&C Division, to simulate
the operation of graphite-moderated, gas-cooled reactors. By adapting
this distributed, object-oriented Java interface to physiological modeling,
Strickler implemented a complex model consisting of many differential
equations in a form that can be remotely controlled by the user. For
example, it shows the effects that a slight change in the frequency
of the heartbeat can have on cardiovascular functions, such as blood
pressure.
|
NetSolve allows the client at the desktop computer to tap into supercomputing levels of power by finding appropriate computational resources, including supercomputers networked into a computational grid, and making them easily accessible to the individual user.
|
One computer science development that will
be useful is NetSolve, a client-server system that enables desktop computer
users to solve complex scientific problems by tapping into supercomputing
levels of power. NetSolve was developed under the leadership of Jack
Dongarra, an ORNL-University of Tennessee Distinguished Scientist. "NetSolve,"
says Ward, "couples desktop computer software to models and problem-solving
programs on a high-performance computer such as the IBM SP at ORNL."
Dongarra and his collaborators are adapting NetSolve for use in the
Virtual Human Portal. David Walker of CSMD is developing abstract concepts
for the problem-solving environment (PSE) of the Virtual Human Portal.
Walker, Dongarra, and Wardalong with researchers from the University
of Utah, Indiana University, and the University of Washingtonare making
an information technology research proposal to the NSF for funds to
develop the computational infrastructure for the Virtual Human Project.
"This infrastructure will include a collaborative environment for model
integration," Walker says, "It will permit collaborators from around
the country to construct integrated models for simulating the structure
and function of the human body. These models will then be executed on
distributed computing resources using sophisticated scheduling mechanics
that make the whole process transparent to the user."
If funded, the
team would develop a generic advanced, distributed PSE, which is an
integrated computing environment for composing, compiling, and running
applications in a specific problem area. Created from the generic PSE,
a prototype Virtual Human will facilitate developing, integrating, and
solving complex 3D models of human physiology and anatomy.
"We envision
developing a smart environment with intelligent agents on a remote server
that would be given a problem to solve," Ward says. "The agents would
find the appropriate set of models and execute them on a computational
grid of high-performance computers. We will also develop a user interface
to present visualization of the results of the simulation either to
the client's desktop or to a room-sized immersive visualization environment
called a CAVE."
What It Means to You
The beauty of a computer model
of a human is that it can be customized for a specific person at any
point in time. At least, that's the long-term vision shared by Easterly
and his colleagues. Customization is important because of evidence that
men, women, and children respond differently to various drugs, drug
dosages, and other treatments, as well as environmental insults.
"From
our earlier work in modeling children's organs," Ward says, "we see
the need to build human models for different ages, sizes, and sexes.
By using equations and changing some parameters, we can make the heart
smaller or larger. We can make a human model or phantom grow with age."
Use of a customized modela computerized clone of you that includes
your genetic makeupwill make it possible to predict how you might respond
to different doses of radiation, chemicals, and drugs, or what damage
you might suffer if you were in an automobile accident or airplane crash.
|
The initial and final, or deformed, shapes of a finite-element mesh model of the human body based on Visible Human data are shown here. The deformation resulted from a wood block moving at constant velocity into the chest just above the heart. The elapsed time for the simulation is 0.0001 second. The spine, heart, lungs (not shown in this view), and surrounding tissue are each assigned appropriate measures of density and elasticity. Simulations such as these can model the impacts on the body of seat belts, air bags, or objects in an automobile during an accident.
|
In fact, one of the early applications of
the Virtual Human model will be the prediction of responses of the body
to impacts from, say, rubber bullets. ORNL researchers doing this research
can draw on some of the Laboratory's work in developing detailed models
of cars and using them to predict the damage these cars will sustain
in collisions at different velocities. In this case, the finite element
mesh software used for cars to pinpoint the effects of pressure on material
deformation will be applied to meshing the human model. This type of
model can also be used to predict the tissue damage that will result
from various types of car wrecks and plane crashes.
"A Virtual Human
model of the lung could be useful for virtual surgerythat is, to simulate
the effect of lung surgery on a patient's respiration before any surgery
is actually performed," Ward says. "A model of the circulatory system
could help doctors decide where to replace part of an artery to reroute
the blood so it flows around a blockage. A model of a patient's hip
could help a surgeon determine the effect of a hip implant on the joint's
movement."
Looking ahead five to ten years, even decades from now, Easterly
sees a "human model on a chip," or at least mounds of personal genetic
and other medical data squeezed into a plastic 3 x 5-in. card. When
plugged into a handheld computer, your card might provide you and your
doctor the results of your X-rays and CAT and MRI scans, your responses
to past treatments, your genetic makeup, and a list of foods and drugs
you should avoid. "The more you know about your body," Easterly says,
"the more you can take charge of your well-being."
The idea of the Virtual
Human Project is slowly creeping through the scientific community. If
the vision continues to spread, it may someday bear fruit of value to
virtually every human.
Beginning
of Article