New Algorithm Lets Your Phone Track Your Wheezes
(Inside Science) -- A simple breath can reveal a lot about someone's health. That's why researchers are developing wearable devices that keep continuous tabs on the wearer's lungs. But while gadgetry often gets the attention, it's the algorithms behind the gadgets that tell you what the data means.
Now, researchers have developed a new algorithm that detects and measures wheezing, providing crucial information to help doctors monitor and understand diseases such as asthma, lung cancer, cystic fibrosis and chronic obstructive pulmonary disease -- or COPD.
The new method is not only more accurate than previous algorithms, according to the researchers, it also requires less data and computer power to extract the information a doctor needs to know about a patient's breathing.
"Our approach outperforms anything that we know out there in the literature," said Hamid Krim, an electrical engineer at North Carolina State University in Raleigh, who developed the algorithm with Saba Emrani, a graduate student at N.C. State. They're presenting the work at the 2015 European Signal Processing Conference in Nice, France, on September 3.
Although the algorithm can be used with all kinds of devices, the researchers envision that it will be part of a system connecting a smartphone with wireless sensors worn on the chest that track every breath an individual takes. Even through background noise, the algorithm can pick out when a patient is wheezing, and identify the characteristics of those wheezes.
While a healthy breath is chaotic, containing no structure in its sound, a wheeze consists of numerous tones at different intensities, varying over time. To detect wheezing, the algorithm looks for those periodic and repetitive patterns.
"A wheeze is like a whistle in the lungs," Krim said. "We take the signal and kind of bend it on itself and see whether there is some structure in the way things are repeating or trending over short times."
By measuring how a person is wheezing, a doctor can figure out what's restricting airflow and better understand a respiratory disease. In general, higher frequency sounds mean something's obstructing the smaller airways in the lungs while lower frequencies suggest blockage in the larger airways. Wheezing is also an early indicator of lung cancer, and the algorithm can distinguish a tumor-caused wheeze from one due to asthma.
But the most useful application might be for monitoring people with asthma and COPD, said Eric Larson, an engineer at Southern Methodist University in Dallas, who's not involved with the new work. According to the World Health Organization's latest estimates, 235 million people around the world suffer from asthma, and 64 million have COPD, which encompasses a variety of maladies that restrict breathing. In 2012, COPD killed 3 million people, and by 2030, WHO predicts that it will be the third leading cause of death worldwide.
But because researchers haven't yet had the technology to continuously monitor breathing, they don't have a baseline to fully understand what breathing difficulties mean for health, said Larson, who helped develop a smartphone app called SpiroSmart, which also monitors lung health. By tracking patients, doctors can learn how people manage the disease, how it worsens or improves over time, and whether certain triggers in the environment exacerbate it.
The new approach is a signal-processing algorithm, which can be adapted to analyze signals other than the sound of wheezing. Ultimately, Krim wants to use the method to monitor all kinds of health indicators, from electrocardiogram readings to physical movements. Paired with an accelerometer, it might be able to analyze a person's gait, distinguishing a walk from a run. In fact, Krim is part of the Center for Advanced Self-Powered Systems of Integrated Sensors and Technologies, an effort involving several universities to create a suite of self-powered sensors to monitor the body and environment, giving a real-time, holistic picture of health.
In the case of asthma, for example, such a system might detect a combination of wheezing, certain particulates in the air, and other physiological signs and warn an asthma sufferer of an impending attack, said Alper Bozkurt, an engineer at N.C. State who's testing the sensors and the algorithm.
Although the fact that the algorithm works with limited data is a huge advantage, it's too early to tell how much of an impact the new method will have, Larson said.
"Most likely this is a nice incremental advance for right now." Configuring the algorithm to work with smartphones can be difficult. And, the researchers still need to test their algorithm on real-life humans in real-life situations.
Eventually, though, it could help people manage diseases like asthma, which often require treatments that differ from person to person.
"Being able to understand how well something is working for one individual is a huge cost-cutting method," Larson said. "And it could improve outcomes and possibly save lives."