Respiratory diseases, such as asthma and chronic obstructive pulmonary disease, impose a significant burden on patients and society as a whole. These conditions can severely impact a person's quality of life, leading to difficulty breathing, fatigue, and decreased physical activity. They can also lead to hospitalizations, missed work and school days, and even premature mortality.

According to the World Health Organization, respiratory diseases are a leading cause of death globally, accounting for approximately 3.9 million deaths each year. In the United States, respiratory diseases are the third leading cause of death, and the prevalence of these conditions is expected to increase in the coming years due to factors such as aging populations and rising rates of obesity.

Effective prevention and treatment of respiratory diseases is crucial in reducing the burden on patients and society. As the old proverb goes, an ounce of prevention is worth a pound of cure, and that certainly holds true when it comes to respiratory diseases. Detecting a latent problem before it becomes a major source of morbidity is the ideal scenario as it allows for early treatments to be initiated, and often results in better long-term patient outcomes.

Unfortunately, early detection of an as yet asymptomatic respiratory disease is a difficult task. Detecting problems at an early stage requires continuous monitoring of breathing patterns over an extended period of time, however, most instruments that analyze breathing are cumbersome and must be operated in a clinic by trained healthcare professionals. As you might expect given these constraints, individuals are rarely put under observation until after significant symptoms have already presented.

Researchers at the National Institute of Standards and Technology in Boulder, Colorado have breathed fresh air into the field with their invention called BreatheSmart. Their technique coopts the Wi-Fi signals that are already bouncing around virtually every indoor space these days to unobtrusively monitor one’s breathing patterns passively and contactlessly. By intercepting certain signals that are sent between client devices and a router, then using a clever machine learning-based algorithm to interpret those signals, the team has shown that it is possible to recognize anomalous breathing patterns.

Small channel state information (CSI) messages are regularly sent from client devices to Wi-Fi access points to help them optimize the transmission link. An access point knows what a CSI message should look like, but as it travels through the environment it gets distorted and loses strength, and these deviations from the true signal provide the information needed to optimize transmissions. These are the messages that the researchers leveraged to monitor the breathing of individuals.

As CSI messages travel between devices, they bounce around the environment and are changed by the objects that they reflect off of, including humans. The team modified router firmware to request these CSI messages more frequently, up to ten times per second, to increase the resolution of the data. By capturing these signals, they could then see how they were changed after interacting with a person in the area. But, determining how these signal modulations corresponded with specific breathing patterns was by no means obvious.

So, to interpret the signals, a deep learning classifier was developed. Such classifiers excel at ingesting raw data and finding complex patterns, making this tool ideally suited to this task. After training the classifier, a series of experiments were conducted using a breathing manikin that can simulate various respiratory disorders in an anechoic chamber. Under these conditions, BreatheSmart was found to be capable of identifying certain specific breathing patterns accurately in more than 99% of cases.

While this work is very promising, there is still much yet to be done. It has not yet been shown how BreatheSmart operates under more natural conditions, outside of an anechoic chamber, for example. It is also unclear how to deal with situations in which the subject being monitored is moving about normally, as opposed to staying motionless aside from breathing, as is the case with the manikin. The researchers will also need to investigate how they can adapt the system to work when multiple people are present in the area being monitored. The team is currently hard at work on these problems, and more, to try to build a practical system for real-world use.

Source link