Building a Supermodel

In search of a versatile virtual person

More than 1,800 years ago, the Greek physician Galen came up with a model for how the human body works. Blood, which he called the “natural spirit,” originated in the liver and traveled to all parts of the body through the veins. In the left side of the heart, some of the blood mixed with air from the lungs to make “vital spirit,” which flowed to all parts of the body through the arteries. Some of this fluid went to the head, where it was infused with “animal spirit,” or consciousness, and then was distributed to the muscles and the senses through a network of hollow nerves.

An internal view of the Virtual Lung, which represents the complicated airway system as a series of tubular Y’s. T. Martonen, J.D. Schroeter, et al/Env. Protection Agency.

Jack, a simulation that can be scaled to represent a person of any size, helps equipment makers design cabs with adequate visibility and control levers that are easy to reach. Engineering Animation, Inc.

Galen was wrong, of course, but his ideas held sway for more than 1,400 years for one simple reason: His model provided a rational explanation of all the biological facts known at the time. Not until the Renaissance rekindled the spirit of scientific inquiry did new observations by Leonardo da Vinci, Andreas Vesalius, and others pave the way for William Harvey’s model of circulation, which appeared in 1628.

Today, the burgeoning volume of biological information has grown so large that it threatens to overwhelm rather than enlighten scientists. With the power of the computer, however, researchers are using the available raw data to refine the crude qualitative models of yesteryear into finely tuned simulations that can represent—and even predict—the behavior of complex biological systems.

Individually, these simulations provide scientists with the opportunity to understand aspects of human biology that aren’t readily apparent and to think about old problems in new ways. Integrated into a whole, they promise a sophisticated version of Galen’s model—a virtual person that could serve myriad roles, such as a computerized crash-test dummy or perhaps as a tool for predicting adverse drug interactions.

Inhaled particles

Over a decade ago, Ted Martonen envisioned developing a computer model to predict how inhaled particles, such as soot, dust, and cigarette smoke, would spread through the respiratory system. Those predictions, he believed, could help physicians target drugs to treat the ailments that those particles cause.

First, Martonen’s research team at the Environmental Protection Agency’s National Health and Environmental Effects Research Laboratory in Research Triangle Park, N.C., developed a mathematical model of the structure of the human respiratory system. The system and the model include the nose, mouth, throat, sinuses, and 20 million airways in the lungs.

Team members then used computational fluid dynamics—the same techniques aeronautical engineers use to analyze the flow of air around aircraft—to determine the airflow through the model. With this information, the researchers could predict how deep the inhaled air pollutants would spread through the lungs and whether they would stick there or be exhaled.

The resulting supercomputer simulation earned Martonen the 1997 Computerworld Smithsonian Award for Medicine. Called the Virtual Lung, the model can be used to help evaluate a patient’s condition and select the best drug and delivery method, the researchers say. The Virtual Lung is now being tested in clinical trials in Canada and the United Kingdom, where it’s being used to customize the delivery of drugs in patients with respiratory-tract diseases, such as asthma and emphysema.

“I wanted to use computers to do what they do best, which is to run thousands of possible [treatment options] to customize drug-delivery protocols in real time,” Martonen says. “I’d like to see the day when a doctor has a stethoscope in one pocket and a floppy disk with our program on it in the other.”

Faster processing speeds

As the processing speed of computers increases, programmers can broaden the potential applications of their models by refining existing capabilities and adding new ones. Nothing illustrates this phenomenon better than Jack, a program that helps engineers assess workplaces according to issues such as ergonomics and potential for causing injury.

Developed at the University of Pennsylvania’s Center for Human Modeling and Simulation, Jack began as a simple computer model that could be scaled to represent a person of any size and weight. Over the years, programmers endowed Jack with the ability to walk, lift objects, and avoid obstacles. The simulation became so successful that it turned into a commercial venture. It is currently owned by Engineering Animation, Inc., in Ann Arbor, Michigan.

Even though Jack is just a teenager, he has quite a résumé. Soldier, pilot, astronaut, assembly line worker, truck driver, and yard-maintenance worker are just a few of his former job titles.

Heavy-equipment manufacturer Deere & Co. has used Jack over the past 10 years, says Jerry Duncan, senior staff engineer in the company’s product technology group in Moline, Ill. Jack has worked for Deere on a virtual assembly line to help equipment designers optimize the size, shape, and location of access panels. He has also operated virtual earthmoving equipment to ensure there’s adequate visibility from the vehicle cab.

“Jack helps us evaluate a wider range of equipment features and the placement of controls without ever putting a human being into the situation,” Duncan says. “We can actually see what it’s like to reach over and pull a lever, for a small female versus a large male, without ever building a prototype.”

NASA uses Jack to simulate a person in surroundings as diverse as the shirt-sleeve environment within a spacecraft and the cramped conditions inside a space suit during extravehicular activity. Recently, Jack helped designers of the International Space Station determine where they should place the footholds for astronauts performing assembly tasks in the weightlessness of space.

Jack is a very reliable worker, NASA has found. About 94 percent of the results simulated with Jack proved acceptable when verified with tests in neutral-buoyancy tanks, said Charles Dischinger Jr., human-factors engineer at NASA’s Marshall Space Flight Center in Huntsville, Ala.

Human biology

Imagine, if you will, a Super-Jack—not one that leaps tall buildings in a single bound but a simulation of the complete human biology. It would encompass full-scale anatomy and continue down to the genetic code.

Only recently have advances in computational tools and biological insights come together to make this type of simulation feasible. Researchers at Oak Ridge (Tenn.) National Laboratory have begun a project to create just such virtual humans, says Clay Easterly, the lab’s director for the Virtual Human project.

The scientists held meetings at the National Academy of Sciences in Washington, D.C., in October 1999 with representatives of several agencies, including the Department of Defense, the National Institutes of Health, and the National Science Foundation. The researchers hoped that interested agencies would commission a study to be presented to Congress that would outline the feasibility, cost, and scope of the project and suggest the approach and who should be involved.

In the meantime, more than 70 academic, government, and corporate scientists met early last November in Washington to begin charting a roadmap for what will undoubtedly be a decades-long effort.

“The Virtual Human will be a computational model of an individual that contains the knowledge base to describe mental health to the extent possible, disease processes, responses to chemicals in the environment, responses to stimuli, and a host of other biological phenomena,” Easterly says.

Early versions of the Virtual Human would focus on integrating existing models of human biology with a detailed model of anatomy. It would begin with the National Institutes of Health’s Visible Human Project, based in Bethesda, Md., which has generated complete three-dimensional digital representations of the normal male and female human bodies with resolution as small as one-third of a millimeter.

Subsequent versions of Oak Ridge’s Virtual Human would incorporate new models as they are developed and refine existing simulations as more data are gathered.

Easterly regards the Virtual Human as more than just a computerized person. It would be a tool to help scientists work with the vast amount of information on human biology that scientists have amassed. This will be quite a chore—the National Library of Medicine’s MEDLINE database has more than 10 million references to health-science journal articles published since 1966, and it expands by about 7,300 references each week.

“We have journals full of specialized data that are understood in large part only by people within that specialty,” Easterly says. “The Virtual Human can help organize this vast amount of information in such a way that a bigger section of the scientific community and the public can appreciate the knowledge, and it will become apparent how all of the pieces affect the whole.”

For example, data from a simulation of the effects of pollutants on individual lung cells could be combined with a model of lung function to help physicians understand how pollution exacerbates congestive heart failure and other conditions.

Passing information

Because the individual simulations that might be assembled into an early version of the Virtual Human have been programmed to run on computers as disparate as Compaqs and Crays, researchers need to develop effective ways to pass information quickly back and forth between the models.

The best way to integrate the models may be to have a distributed network of computers run the Virtual Human simulation—think of it as living with your kidneys in Cleveland and your immune system in El Paso. Distributed computing would provide many beneficial side effects, says Michael Ackerman, assistant director of the National Library of Medicine’s Office of High Performance Computing and Communications in Bethesda, Md.

“In a networked type of model, the individual models will stay in the hands of those who developed them,” Ackerman says. “Each expert provides his own piece of the program and is responsible for its upkeep and maintenance, and the model resides wherever the researcher is.

That way, the simulation can be updated to include the latest and greatest information available.”

The possible uses for a high-fidelity simulation of human biology are innumerable, and the benefits of constructing such a model are likely to be wide-ranging, Easterly says. The Virtual Human could, for example, alert doctors to possible harmful interactions between drugs, foods, and nutritional supplements.

Also, the blunt-trauma models needed to evaluate nonlethal weapons for the Department of Defense could be used to design better safety equipment, such as bulletproof vests, motorcycle helmets, and seat belts. Such computer simulations would offer considerable savings, compared with traditional evaluation methods.

“The Virtual Human is likely to be of enormous value because it will be an environment in which we can test hypotheses and evaluate strategies for treatment in the context of what we think we know about human biology in silico before we ever try them out in patients,” says Tim Buchman, chief of surgical critical care at Barnes-Jewish Hospital in St. Louis.

Says George Fenton, who heads the Joint Non-Lethal Weapons Directorate at Quantico (Va.) Marine Corps Base, “The Virtual Human project is a great idea. The Departments of Transportation, Defense, Justice, and Energy, as well as other organizations, will all be working toward a common goal, and we’re looking to leverage the power of the computer to help us solve these problems.” Fenton adds, “I think it’s going to happen sooner than most of us think.”

Personalizing simulations

Although in the near term the Virtual Human would be a generic human being, the long-term goal is to be able to personalize the simulation by superimposing an individual’s physical characteristics, medical history, genetic information, and other personal data on the general model. Such a model could be used to assess the best course of treatment and tailor a patient’s prescription dosage to meet individual needs.

Richard M. Satava, a surgeon at Yale University School of Medicine in New Haven, Conn., sees the personalized Virtual Human as an integral part of a suite of new technologies for the doctor’s office of the future.

The patient would enter the office through a portal, similar to an airport metal detector, that conducts a scan of his or her structural anatomy, Satava says. Next, he or she would proceed to a waiting room where technicians would place the patient’s palm onto a table-top scanner to gather vital signs and other medical data. All this information would be instantly uploaded to an individualized version of the Virtual Human, which would serve as a next-generation medical record and fit on a memory cartridge the size of a credit card.

When the doctor slips the cartridge into a holographic viewer, a three-dimensional image, annotated with pertinent medical data, would pop up. If the doctor suspects a liver problem, for example, he or she could enlarge the image and move through and around the virtual organ to conduct a more detailed investigation.

If the problem requires surgery, the images could be used for preoperative planning or downloaded into a surgical simulator that would enable doctors to practice the operation before they pick up a scalpel.

All this fantastic technology is years in the future, of course, but Satava sees the Virtual Human as a springboard for its development.

“This project would be the health-care equivalent of the Apollo missions to the moon,” Satava says. “It could be the pivot point for a new information revolution in the health-care system that would dwarf what we’re seeing today with the Internet.”

More Stories from Science News on Tech

From the Nature Index

Paid Content