Tweet This
As we know by now, Alexa can play a song, order a pizza or do a quick online search.
But now it can do something much more valuable: save your life.
According to the results of a new proof-of-concept study, Alexa can accurately identify a specific pattern of breathing known as agonal breathing or gasping for air , that develops in the setting of an impending cardiac arrest, or when your heart stops beating.
The research was published yesterday in the npj Digital Medicine.
The implications for this novel form of contactless AI monitoring to detect cardiac arrest are broad, and offer the unique possibility to dispatch an ambulance to a victim who may be alone at home.
Current data indicates that close to 500,000 people in the U.S. die annually from cardiac arrest . And one of the most common areas where this technology may be most valuable is in the bedroom, where a large percentage of out of hospital cardiac arrests occur.
Immediate CPR can double or triple someone's chance of survival, but requires a bystander to be present to perform chest compressions.
The Study
Researchers at the University of Washington have developed technology to monitor people for impending cardiac arrest as they sleep, without the use of any patch, attached leads or body monitor—just the audible pattern of sounds of breathing as they sleep.
It involves the ability of a smart speaker, such as Amazon Alexa, Google Home or even a smartphone speaker to detect the gasping sound of agonal breathing and then dispatch 911 for help.
The technology, which was developed from monitoring and analyzing actual episodes of agonal breathing from 911 calls, was able to detect agonal breathing events 97% of the time from up to 20 feet (6 meters) away.
"A lot of people have smart speakers in their homes, and these devices have amazing capabilities that we can take advantage of," said corresponding author Shyam Gollakota, an associate professor in the UW's Paul G. Allen School of Computer Science & Engineering, in a press release. "We envision a contactless system that works by continuously and passively monitoring the bedroom for an agonal breathing event, and alerts anyone nearby to come provide CPR. And then if there's no response, the device can automatically call 911."
Based on review of 911 call data, researchers noted that agonal breathing occurs in approximately 50% of persons who experience cardiac arrest. Agonal breathing is reflective of low oxygen levels in the body, known as hypoxia. Research also suggests that those persons who take agonal breaths typically have a better chance of survival.
"This kind of breathing happens when a patient experiences really low oxygen levels," said another co-author of the study, Dr. Jacob Sunshine, an assistant professor of anesthesiology and pain medicine at the UW School of Medicine. "It's sort of a guttural gasping noise, and its uniqueness makes it a good audio biomarker to use to identify if someone is experiencing a cardiac arrest."
The researchers collected sounds of agonal breathing from actual 911 calls to Seattle's Emergency Medical Services (EMS) from 2009-2017. In the setting of a cardiac arrest (with people unconscious), bystanders used their phones to record the agonal breathing sounds, placing their phones close to the patient's mouth and face so that the dispatcher could figure out whether the patient needed immediate CPR.
The team collected 162 calls between 2009 and 2017 and extracted 2.5 seconds of audio at the start of each agonal breath to come up with a total of 236 clips. The team captured the recordings on different smart devices: an Amazon Alexa, an iPhone 5s and a Samsung Galaxy S4. Additionally, they also used various machine learning (ML) techniques to increase the amount of data to 7,316 positive clips.
"We played these examples at different distances to simulate what it would sound like if it the patient was at different places in the bedroom," said lead author Justin Chan, a doctoral student in the Allen School. "We also added different interfering sounds such as sounds of cats and dogs, cars honking, air conditioning, things that you might normally hear in a home."
For the so-called “negative” data--referring to typical or expected sounds associated with sleep--the researchers used 83 hours of audio collected during sleep studies resulting in 7,305 samples of ambient sounds. These clips included common sounds that people make in their sleep, such as snoring or even periods of pauses in breathing that occur in people with obstructive sleep apnea.
From these datasets, the team used ML techniques to develop an application which could detect and differentiate agonal breathing 97% of the time when the smart device was placed up to 6 meters away from a speaker generating the sounds.
The researchers then tested the algorithm to be assured that it wouldn't mistakenly identify a different type of breathing such as snoring, as agonal breathing. Calling 911 or family members when a patient is just sleeping or napping certainly wouldn’t be ideal.
"We don't want to alert either emergency services or loved ones unnecessarily, so it's important that we reduce our false positive rate," offered Chan.
For the sleep lab data, the false positive rate-incorrectly identifying a breathing sound as agonal breathing when it was actually normal--was quite low, 0.14%. The false positive rate for separate audio clips was 0.22% for separate audio clips, in which volunteers had recorded themselves while sleeping in their own homes. However, when the team tweaked the algorithm, evaluating its ability to detect two distinct events at least 10 seconds apart, the false positive rate fell to 0%.
However, the greater value of this technology may actually be for patients with obstructive sleep apnea (OSA), a disorder in which typically obese patients experience periodic pauses in their breathing, ultimately related to excessive relaxation of enlarged muscle and soft tissue in the throat and upper airway during sleep. With this disorder, patients may experience elevated levels of carbon dioxide and a drop in oxygen saturation, which can lead to stress on the heart, elevated pressure in blood vessels in the lung, and respiratory compromise.
The researchers believe that this algorithm could potentially be developed into an app on a smartphone while people sleep, or a program that Alexa could run as a background monitor on a smart speaker.
"This could run locally on the processors contained in the Alexa. It's running in real time, so you don't need to store anything or send anything to the cloud," Gollakota said.
While this may be true on a local level, without any cloud involvement, the issue of privacy and the ability to prevent any data intrusions is critical. The ability to engage such technology shouldn’t mean that your privacy could be breached, especially at such a critical moment.
"Right now, this is a good proof of concept using the 911 calls in the Seattle metropolitan area," he said. "But we need to get access to more 911 calls related to cardiac arrest so that we can improve the accuracy of the algorithm further and ensure that it generalizes across a larger population."
"Cardiac arrests are a very common way for people to die, and right now many of them can go unwitnessed," Sunshine said. "Part of what makes this technology so compelling is that it could help us catch more patients in time for them to be treated."
But one expert, Satjit Bhusri, M.D, FACC, Associate Program Director, Assistant Professor of Cardiology, Lenox Hill Hospital, Northwell Health, argues that by the time agonal breathing develops, it may be too late to make a meaningful difference, especially in terms of the value of CPR and ongoing resuscitation.
"Agonal breathing is always a sign of end stage disease, whether cardiac arrest or otherwise. It is our goal to catch patients before they reach this stage," he explained.
"Detecting this in people as a measure of cardiac arrest essentially suggests that the patients is extremely critical to terminal. As such many ethical issues are brought up with this diagnosis. End of life care, heal care proxy, etc. are all very private and guarded personal decisions. Privacy would be a major concern when it is unknown what that patient's goals of care are and to whom they would like to know about their condition."
Another expert also expressed caution regarding the use and value of such technology, especially at this early stage
"I think it's clever and represents yet another way to track someone at risk [for cardiac arrest], but it is creepy to have continuous listening surveillance and impractical--we'll see how it plays out," said Eric Topol, M.D., an expert in AI, Editor-in-Chief of Medscape and Director, Scripps Translational Science Institute, Executive Vice President and Professor, Molecular Medicine, Scripps Research. "The concept may be more useful for sleep apnea diagnosis and monitoring as an adjunct to oxygen saturation."
Dr. Topol's newest book, Deep Medicine, explores the intersection of AI and heathcare, and how we can use this valuable technology in the best interests of patients and healthcare professionals alike.
