Making the Greatest ImpactThis story is part of an ongoing series illustrating how the university — through its community, its creative culture, its unmatched scope and scale — is having and will continue to have a greater impact than any institution of its kind on issues, people and events that are changing the world. Read more about how we make the greatest impact. >

Wireless worries

With the help of supercomputing power, researchers explore the health risks of wireless devices through virtual body models and advanced algorithms

Oct. 17, 2011

video icon VIDEO

Every moment, we’re swimming in a sea of electromagnetic radiation. Appliances, power lines, cellphones, Wi-Fi and a slew of other modern technologies emit microwaves that pass through, and interact with, our bodies.

As wireless technology continues to proliferate in our daily lives, anxiety builds about its dangers.  Do cellphones cause cancer? Impact fertility? Affect pregnancies? Increase brain activity?

Yilmaz and his team
An original axial MR image from the U.S. National Library of Medicine’s Visible Human Project® from which Yilmaz and his team created their computer model.

video icon Watch a video about the AustinMan project featuring the research of Yilmaz, his colleagues and his students.

In 2009 the National Science Foundation (NSF) funded a five-year interdisciplinary study at The University of Texas at Austin to address the growing debate about the effects of microwave radiation. After two years, Electrical & Computer Engineering Assistant Professor Ali Yilmaz and his colleagues have built one of the highest-resolution electromagnetic human models to date: AustinMan. The model is helping to determine the effects of microwaves from wireless devices on the body.

“You have all these bright future pictures of connectedness and then all these scare stories telling you to turn off your cellphones,” said Yilmaz. “What are the effects of all these nearby wireless devices on us?”

Yilmaz has worked with his colleagues in the Electrical and Computer Engineering, Aerospace Engineering and Engineering Mechanics and Computer Science departments, the School of Biological Sciences and the Texas Advanced Computing Center (TACC) to answer this question.

Their approach is different than those based on observational epidemiology, which uses statistical correlations to estimate the risks of wireless devices — for example, by comparing the rate of malignant brain cancer in sample populations with the total number of hours they spend talking on the phone.

Based on these studies, in May 2011, the World Health Organization (WHO) released a tentative warning (PDF), suggesting pragmatic measures to reduce exposure when using radiofrequency electromagnetic power-emitting devices near the body.

Ali Yilmaz with students Max Wei and Jackson Massey
Professor Ali Yilmaz discusses the AustinMan project in the ACES Visualization Lab with his students Max Wei and Jackson Massey.

“The evidence, while still accumulating, is strong enough to support a conclusion… that there could be some risk,” wrote Dr. Jonathan Samet, chairman of the WHO’s working group on the subject. “Therefore, we need to keep a close watch for a link between cellphones and cancer risk.”

According to Yilmaz, one should be careful about interpreting these kinds of data.

“Epidemiological studies are limited by many different types of biases and flaws and necessarily trail the technology,” Yilmaz said. “They can discover correlation but not causation.”

In general, a strong correlation between an environmental agent and an observed effect indicates that the first is likely causing the second. However, when there are no biologically plausible mechanisms by which exposure to the agent could cause the observed outcome, the correlation must be especially strong for scientists to reach that conclusion.

In the case of non-ionizing electromagnetic fields such as those produced by wireless devices, no plausible mechanisms for cancer have been established yet. What are established, however, are the thermal, or heat-based, effects of radiation.

Thermal Damage and Safety Standards

Microwaves, when absorbed by our bodies, heat our cells.

Although mechanisms such as blood flow and sweating can delay or reduce temperature increases, if the microwave power is high enough, it can cook tissue like meat in a microwave oven.  Even at much lower power levels, tissue damage and adverse health effects have been observed. In animals, the effects range from fetal malformations to changes in the permeability of capillaries in the brain.

AustinMan displayed on Stallion
AustinMan was created from high-resolution scans of a human body. Supercomputers at TACC make it possible to provide a great amount of detail of different layers, from tissue to bones to blood vessels. Simulations are displayed on Stallion, the world’s highest resolution tiled display.

To protect human beings against these thermal effects, regulatory bodies have issued safety standards that restrict our exposure to microwave fields. Since 1996, the Federal Communications Commission has required that any wireless device sold in the United States comply with these standards.

“The standards require that the microwave power absorbed by our bodies due to a wireless device, quantified in terms of the specific absorption rate, be smaller than the specified limit to keep thermal damage at bay,” said Yilmaz. “The problem is, we don’t know precisely how much power is absorbed.”

Even when the power supplied to the device is known, the absorbed power varies from person to person and depends on factors such as the person’s size, posture and water content.

Because of the difficulty of measuring the power absorbed by different people under different conditions of use, device manufactures use something called a Standard Anthropomorphic Model (SAM, for short) to demonstrate compliance with safety standards. The crash test dummy of the cellphone world, SAM experiments leave much to be desired.

“It’s essentially a head-shaped bowl with some water in it, some salt, some liquid that is supposed to represent your brain,” said Leszek Demkowicz, professor of aerospace engineering and engineering mechanics at The University of Texas at Austin. “It doesn’t really capture any details or any heterogeneities. They put a temperature probe inside the liquid and put a cellphone next to SAM and say, ‘How much heating is there? Does it satisfy the regulatory limits?’ That’s the industry standard. If SAM is cold enough, your device is compliant.”

Simulation-based Studies

Yilmaz and his colleagues at the Institute for Computational Engineering and Sciences endorse an alternative way to study the heating effects of wireless devices — using computer models and algorithms to estimate the absorbed power.

Chart showing the range of both ionizing and non-ionizing radiation
Chart showing the range of both ionizing and non-ionizing radiation. Radiation from wireless devices isn’t the same as that from a nuclear power plant, but the health risks are still unknown.

This approach has been around for decades, but the predictive power of older computations was limited by the complexity of the problem. “Even the latest generations of supercomputers are challenged by these simulations,” said Yilmaz.

Simulation-based power absorption studies face three roadblocks. First, high-precision human models are needed to ensure the precision of the results. Second, high-accuracy algorithms are needed to quantify and limit computational errors in the simulations. And, third, highly scalable parallel software is needed to take advantage of cutting-edge supercomputers.

Yilmaz and his colleagues have been addressing each of these roadblocks. They have recently created AustinMan, a publicly available model that represents the human body with one-millimeter-cubed resolution (something akin to a virtual Lego body composed of extremely small parts). It is being used to perform simulations that predict the electromagnetic power absorbed by our bodies.

The model has an interesting backstory. In the 1990s, a Texas death row inmate donated his body to science; it was used to create ultra-high-resolution scans of the body’s tissues.

“The resolution of these images is higher than what you can get with even the latest MRI and CAT scans,” Yilmaz explained. “We can even see the damage to the tissues that the lethal injection caused.”

To create the AustinMan model, the group worked with anatomists to transform the image slices into computational maps of the body’s tissues. Whereas previous models had included only a handful of tissue types, the current model contains 30 types of tissues, each with unique electromagnetic properties. Overall, the model contains more than 100 million voxels (3-D versions of pixels) that interact with one another during the virtual cellphone calls.

High-Resolution Simulations Lead to More Accurate Predictions

Such extreme simulations are impossible using traditional computing methods and software. Even with the efficient algorithms that the researchers are developing, each simulation would take about five years of continuous execution on an ordinary desktop computer. Crunching the numbers on the Ranger supercomputer at TACC on the Pickle Research Campus in North Austin, however, Yilmaz and his team can perform these simulations in less than six hours.

Yilmaz and his students Jackson Massey, Max Wei and Cemil Geyik
Yilmaz and his students (from left to right) Jackson Massey, Max Wei and Cemil Geyik.

“The supercomputing infrastructure at TACC is fundamental to this work. It’s part of the reason why NSF awarded us this project,” Yilmaz said. “The simulations we’re performing on Ranger are some of the biggest and most complicated bioelectromagnetic simulations ever.”

During the past two years, the project has used more than 3 million computing hours on TACC’s supercomputers, the equivalent of 342 years on a single processor.

The team’s initial results with AustinMan show the importance of having high-resolution body models. When different resolution models are used, the total power absorbed by AustinMan due to an antenna near the ear varies less than 1 percent. However, the power absorbed by other tissues depends quite strongly on the model resolution.

In a recently submitted paper, Yilmaz and his students showed that low-resolution models can under- or over-estimate the power absorbed by the skin, the cornea, the cerebrospinal fluid, and brain matter by up to 50 percent — a significant amount. Many more such simulations are under way.

Without the race car driver, you can’t really get the most out of your race car. My students are the race car drivers in this project.

– Professor Ali Yilmaz
Cockrell School of Engineering

“This has been a very challenging project,” said Jackson Massey, a graduate student in Yilmaz’ lab. “A lot of work has been done on it, developing the model from individual image slices (scanned from the death row inmate). We’ve carefully processed about 2,000 slices that have on average about half a million pixels.”

According to Yilmaz, the student researchers — who helped create the model, develop the software algorithms and come up with ways to harness Ranger — are the creative engines behind the project.

“Without the race car driver, you can’t really get the most out of your race car,” he said. “My students are the race car drivers in this project.”

These simulations will not answer the question of whether cellphones are dangerous per se — much about the dynamics of cancer and other adverse health effects are still a mystery to scientists. But they represent one of the best ways to probe and quantify the thermal effects of wireless devices on the human body. They also act as virtual test chambers that allow the design of better antennas and wireless devices that can operate safely near, on, or in the human body.

“Can we increase the radiated power 100 times to get much better video connectivity? How safe is it to do that?” Yilmaz asked. “If we can’t increase the power, then can we design antennas that minimize the power absorbed by our bodies and maximize the power radiated away? We are developing cutting-edge simulation technology that can help answer these questions.”

For more information, contact: Aaron Dubrow, Texas Advanced Computing Center, 512 475 9439.

To view a linked PDF file in this article, you must first download the Acrobat Reader plug-in for your browser.

1 Comment to "With the help of supercomputing power, researchers explore the health risks of wireless devices through virtual body models and advanced algorithms"

1.  Alex said on Oct. 24, 2011

Has AustinMan revealed any results that are comparable to the FCC requirements? Did the FCC overestimated or underestimated the temperature and heat limit allowed to be absorbed by the head and still be safe? Or is the study still going on?