Perfect Exposure

The VIP in George Xu’s radiological engineering laboratory is not the principal investigator himself, nor is it a postdoctoral fellow or a graduate student. And it is not Karl, the life-size torso made of high-tech plastic.

VIP-Man (“VIsible Photographic Man”) is a computer code containing 3 billion voxels (a voxel is a 3-D pixel) of information. He – yes, VIP-Man is male – is a research subject extraordinaire used to study how radiation affects the human body. Applied to such problems as occupational exposure to radiological contamination and unintended effects of radiation therapy, research with VIP-Man will greatly augment our understanding of how electrons, neutrons and protons interact with and cause damage to human tissues.

With several medical doctors in his family, Xu says he has always wanted to combine his interest in physics with medicine. He saw engineering as a way to study applied physics, and came to the United States to earn his doctorate in nuclear engineering (with a focus on health physics) from Texas A&M University, College Station.

Xu, PhD, is a professor in the department of mechanical, aerospace and nuclear engineering, and the department of biomedical engineering at Rensselaer Polytechnic Institute, Troy, N.Y. He and his virtual patient are collaborating with doctors across the country to improve therapies that use radiation.

Xu leads a team of researchers awarded a three-year, $2.1 million grant from the National Institutes of Health (NIH) to develop 3-D virtual patient models that will more accurately compute radiation doses for CT imaging, nuclear medicine and radiation treatment of cancer patients.

The grant is funded by the National Cancer Institute, part of NIH. Additional researchers from Rensselaer Polytechnic Institute, Troy, N.Y.; Vanderbilt University Medical Center, Nashville, Tenn.; the University of Florida, Gainesville; and Massachusetts General Hospital, Boston, are bringing expertise to the multidisciplinary project in computer science, CT imaging, nuclear medicine and proton therapy.

“Dr. Xu’s research aims to better understand the effects of radiation interaction on the human body using virtual patients, thereby enabling radiologists to use safer and more effective doses of radiation to image and treat actual patients,” says Omkaram “Om” Nalamasu, vice president for research at Rensselaer. “His work is an example of the advanced imaging and computational modeling research being conducted at Rensselaer and how it is collaboratively applied to address pressing medical and healthcare problems.”

From Schematics to Real People

With a nod to its early ancestors, including a tank of water and “Karl,” the phantom torso, Xu created VIP-Man using a comprehensive set of images collected from a cadaver. The image collection, known as the “Visible Human Project,” was an ambitious project undertaken by the National Library of Medicine in the late 1980s and was made available to the public in the mid-1990s.

“Here was this huge dataset out there,” says Xu, “and not too many people had clear ideas about the potential engineering applications of this invaluable set of anatomical data.”

But, Xu had a clear idea. As someone who completed his doctoral research using simple geometric shapes to represent the human body in the early 1990s, he knew exactly how to take advantage of a realistic and descriptive dataset. Picture the difference between a preschooler’s stick drawing and Leonardo DaVinci’s Vitruvian Man.

Xu rendered his “aha” moment into a prestigious Faculty Early Career Development Award (CAREER) from the National Science Foundation in 1999 and successfully used the four-year funding to make his virtual patient, painstakingly combining precise organ anatomy with computer codes simulating the movement of radioactive particles through the body.

Although Xu has subsequently received various grants from the NIH, the Department of Energy (DOE) and the nuclear industry, he says, “It was the CAREER award that really validated my original ideas and gave me the freedom and confidence to pursue innovative research.”

For Xu and his two doctoral students, innovation was key, considering they were the first to attempt simulating radiation in a whole-body image set with so many voxels in it.

In addition to a complex geometry, the human body has tissues that differ in density and atomic composition, affecting the travel of radioactive particles. A tank of water was once used to approximate living tissue – considering humans are more than 50 percent water. Next came high-tech plastic, such that Karl is composed of three “tissue-equivalent” materials to model the density and composition of bone, soft tissue and lung.

VIP-Man trumps both methods by accounting for dozens of tissues with regard to density and composition, not to mention precise scale and anatomical shape. In addition, VIP-Man provides critical new insight into tissues, such as skin, gastrointestinal track mucosa, eye lenses and red bone marrow – which are particularly sensitive to radiation, but were too small to be modeled by physical phantoms or previous computer simulations.

Xu is one of the few experts in the world who have successfully combined the Monte Carlo codes (those used to simulate nuclear weapons) to whole-body human models like the VIP-Man. However, VIP-Man still holds the record for having the largest number of voxels ever used for radiation simulations.

After publishing a series of papers on radiation protection of workers using VIP-Man from 2000 to 2002, Xu turned to medical applications. Radiation is employed in a variety of diagnostic and therapeutic procedures including CT scanning, nuclear medicine and radiation treatment.

Accurately calculating doses to different parts of the patient body in each of these procedures has long been difficult, risking either doses that are too high, which cause side effects, or doses that are too low to effectively treat a tumor. But, modeling specific medical modalities and how the radiation interacts with VIP-Man will provide never-before-seen scientific data to optimize the benefit-to-risk ratios of these procedures to patients.

Virtual Patients

In Xu’s development of VIP-Man, he has nurtured collaborations with medical researchers at Vanderbilt University, the University of Florida and Massachusetts General Hospital to use virtual patients to further medical diagnostics and therapeutics. This led NIH to award Xu and his collaborators the necessary funding over three years to study clinical applications and to expand the virtual subject population. Xu has also been invited to serve on a study section for the Biomedical Information Science and Technology Initiative at NIH.

Brian Wang, a doctoral student who graduated in May from Rensselaer, lends some insight into how Xu manages to attract such ambitious collaborations. When attending a national conference, students in Xu’s lab were instructed to meet 20 new people.

For a research career, interacting with the right people can be more important than learning the results of the presentations at a conference, Xu believes. Wang says Xu led by example, illustrating for his students the value of networking.

“It’s his style,” says Wang, which is why he has such “tremendous connections within the scientific community. Otherwise, it’s a failure in attending such a national conference,” says Wang, who has accepted an offer as a clinical medical physics faculty member at Cooper University Hospital in New Jersey.

Wesley Bolch, PE, PhD, professor of nuclear and biomedical engineering at the University of Florida, describes a scientific session at an international conference that Xu organized in Chattanooga, Tenn., giving him credit for bringing together “the world community.”

Xu planned the meeting with Keith Eckerman, PhD, of Oak Ridge National Laboratory, a world authority on radiation dosimetry, by sending personal invitations to researchers as far away as Japan, Korea, China, Australia and Europe. Although they have read one another’s papers on radiation human modeling, “this group of researchers – doing this kind of work – had never really met in one room before,” says Bolch. The meeting was a scientific success as well, a tribute to Xu’s “personal character, initiative and organization skills,” Bolch says.

Bolch was an assistant professor at Texas A&M University when Xu was a doctoral student there in the early 1990s. They have since become collaborators after Bolch went to UF and Xu joined Rensselaer. In 2001, Bolch invited Xu to give a “Frontiers in Biomedical Engineering” research seminar at the University of Florida on tomographic modeling, for which Xu is now recognized as one of the pioneers.

Bolch and Xu are currently collaborating on developing models of children. One of the projects is to study CT imaging, allowing doctors to visualize internal tissues with high resolution.

However, Xu says, “Most hospitals don’t differentiate [the CT procedures] for patient size,” thus pediatric patients may be receiving unnecessarily high radiation doses. And because children have more years of life ahead of them, they may be more likely to develop long-term effects – such as cancer – from X-ray exposure.

Xu and his Florida colleagues have two aims: to develop virtual patients based on children and to study X-ray interactions with the virtual body from different intensities of CT imaging. Bolch organized a workshop during the Society of Nuclear Medicine’s annual meeting in 2005 in Toronto, and he invited Xu to lecture to an audience of mostly physicians.

At Vanderbilt, Xu’s collaborators, radiology professors Michael Stabin, PhD, and Randy Brill, MD, PhD, work in nuclear medicine. Immunoradiotherapy is like a smart bomb for a cancer tumor. Medicine is administered that contains antibodies (hence, “immuno”) that recognize molecules unique to cancer cells and thereby target the delivery of the treatment agent (the radioactivity).

This clinical application is becoming “increasingly more important” says Xu, as more drugs are developed to image and destroy specific types of cancers at the molecular level. A major problem, he says, is not knowing how much of the injected dose gets to the target site.

“Physicians tend to be overly cautious” when injecting radioactive substances, using lower doses to stay safe from overexposure, says Xu. Using data from virtual patients, doctors can better calculate dose and more aggressively and effectively treat cancers in real patients.

Virtual patients will be used at Massachusetts General Hospital by radiation physicist Harald Paganetti, PhD, and radiation oncologist Herman Suit, MD, PhD, to improve proton therapy. In this procedure, a medical accelerator delivers a beam of protons to the target organ.

According to Xu, “The clinical problems we are trying to solve are how radiation goes into the patients, and how it will cause secondary radiobiological effects” that can result from radiation scatter. The objective is to optimize dose and beam direction to target the tumor while minimizing damage to nearby healthy tissues using the advanced procedures.

Xu and his collaborators plan to have more VIPs, varying in gender, age, size and ethnicity, as he and his colleagues create a family of virtual patients. Although VIP-Man was the most detailed model when first created, he is just a single, very tall, very heavy adult male,” says Xu.

“Currently accepted methods in radiation protection and nuclear medicine do not realistically consider patient variations in age and body size, resulting in very large miscalculations in the true radiation dose to the patient,” says Xu. “Our project aims to bring about a paradigm change by creating a realistic patient model library and related computational tools that will facilitate image processing, simulation and radiation dose measurement for various clinical diagnostic and therapeutic procedures.”

To that end, Xu is leading a worldwide consortium. He is creating a Web site as a master depository for virtual patients and applications. Rensselaer computer science faculty Daniel Freedman, PhD, and Chuck Stewart, PhD, are working with Xu to develop advanced software to handle the huge datasets – including such computational tools as image segmentation, 3-D and 4-D visualization and Monte Carlo dose simulations.

The Web site will allow offsite collaborators to share data and compare results, says Xu, but it is also a mechanism “to disseminate all the information, all the data, freely to the research community.”

As someone who has benefited from open access to a detailed anatomical dataset, Xu fully understands the value of making resources available to the public domain.


Editor’s note: This article first appeared in the Rensselaer Research Review, Fall 2005 edition. ©Rensselaer Polytechnic Institute.