Skip Over Navigation Links

Web Exclusives: Systems

Salvaging Signals for Health
By Emily Carlson
Posted February 2, 2009

Just as automakers use recycled steel scraps to build cars or construction companies repurpose tires to lay running tracks, scientists are reusing previously discarded medical data to better understand our complex physiology.

This movie shows what would happen when bacteria multiply in a confined space.

This movie shows what would happen when bacteria multiply in a confined space. Individual bacteria are color-coded to indicate their orientation to the container walls: from blue (perpendicular) to red (parallel).
These time series show the heart rates of four different individuals. The top recording from a young person indicates the fluctuations of a “healthy” heart. Credit: Madalena Costa and Ary Goldberger
Full-screen version (MOV, 6.81 MB)

About 10 years ago, a team of researchers started “recycling” anonymous medical measurements typically tossed after patient exams. These included complete recordings of heart and brain wave activity. At the time, the scientists harnessed a relatively new technology—the Internet—to develop a Web site called PhysioNet External link where others could access this health information along with freely available software to analyze it.

The purpose of the site is to stimulate new work on complex biomedical and physiologic signals, such as heart rate, respiration, brain activity and gait. Studying these signals in depth could help doctors diagnose and treat health conditions like congestive heart failure, sleeping disorders, epilepsy and walking problems.

“What we want is for people to explore data with new eyes and no biases,” says Ary Goldberger, a cardiologist at Beth Israel Deaconess Medical Center/Harvard Medical School who helped create PhysioNet. “We want them to test ideas that have never before been considered.”

To study how well different parts of our bodies function, doctors typically measure the electrical or chemical signals produced by those parts. They use an electrocardiogram (ECG or EKG), for example, to study the rhythms of a heart after cardiac arrest. The doctors take the average of all the signals collected during the exam to assess physiologic function and determine treatment options.

“The focus has been on looking for information in a rather static snapshot, but our heart rates aren’t static—they change second to second,” says Goldberger. “Information about how healthy or sick [a person] might be is hidden in those fluctuations.”

PhysioNet is dedicated to examining those fluctuations. From the site, researchers can access a growing archive of well-characterized digital recordings from both healthy and sick individuals. They also can freely download around 100 open-source software programs that let them explore the medical measurements in innovative ways.

“Usage has been huge,” says Goldberger. According to user statistics from PhysioNet architect and webmaster George Moody, each month an average of 50,000 different visitors come to the site and upload between 3 and 4 terabytes of data—the same amount as about 1 million songs.

Can you tell which heart rate pattern is health?
This figure shows the heart rates of four people. Only Signal B is from a healthy person. The others are from people at high risk of sudden cardiac arrest or stroke. Credit: PhysioNet

Users exploit the data and software to investigate their own basic research or clinical questions. Goldberger says that findings published in more than 400 papers have stemmed from PhysioNet resources and related materials.

“It’s been particularly exciting to see how college and high school students have been able to use PhysioNet for some really creative, original projects,” says Moody, noting that a high school senior used the site’s materials and methods for a project on heart rate variability that won a gold medal at the Canadian National Science Fair.

Whether students or seasoned investigators, scientists also explore questions posed by the PhysioNet community during annual challenges designed to drive the development of new mathematical tools for mining the data. For one challenge, different teams found patterns in heart and respiratory signals that indicated periods of sleep apnea, a common disorder where people briefly stop breathing while asleep. This work led to a number of patents and applications, including new devices to assess sleep quality more cheaply and quickly than current methods.

The 2009 challenge, currently under way, is focused on developing tools for predicting episodes of hypotension, or extremely low blood pressure, among intensive care unit (ICU) patients. The dataset of physiologic signals from some 2,300 ICU patients suggests that those with hypotension are more than twice as likely to die. Tools to identify which patients are at greater risk could improve their care and survival. Results of this year’s challenge will be announced in September.

“What’s so groundbreaking about all of this is that Physionet provides a much needed way for both researchers and health care providers to obtain integrated sources of knowledge and a means to collaborate. That also makes for a terrific teaching resource,” says Jim Anderson of the National Institutes of Health, which funds the project through its National Institute of Biomedical Imaging and Bioengineering and National Institute of General Medical Sciences.

Goldberger says the next stage of PhysioNet involves the creation of an interactive virtual laboratory community. Much like a profile on a social networking site, research teams could create project home pages for storing, sharing and reviewing data. The goal, he says, is to expand the “virtual data universe” of physiologic signals and our knowledge about them even more.

Learn about related research

This page last reviewed on April 22, 2011