Sleep apnea is the world’s most common sleep-breathing disorder, affecting up to one billion adults worldwide. It is characterized by recurrent pauses in breathing lasting over 10 seconds, and is associated with severe health complications such as high blood pressure, coronary artery disease, stroke, and heart failure.
Despite the fact that it affects about one in seven adults, the majority of sleep apnea cases may still remain undiagnosed. The potential severity means there is a pressing need for better diagnosis of this sleep disorder.
The current gold standard in the detection of sleep apnea is polysomnography, a technique that involves the monitoring of various sleep parameters, including heart rate, respiratory rate, breath depth, blood oxygen levels, brain activity, and leg and eye movements, and combines these observations with electroencephalography (EEG) and electrocardiography (ECG) data.
The problem is this involves staying overnight in health clinics and expensive equipment that can be uncomfortable for patients — things that can disrupt normal sleeping patterns. Home sleep apnea monitoring technology is available, but this often relies on measuring nasal breathing using the application of nasal cannulas and, therefore, can also be uncomfortable for wearers.
These things can result in disrupted sleep patterns, missed breathing patterns, and breaks in monitoring, which pose a significant challenge for sleep apnea diagnosis.
A new wearable monitor
New research published in the journal Advanced Intelligent Systems describes a novel sleep apnea detection system in the form of a wearable device that tracks the breathing of patients without the need for sleep clinics and uncomfortable and expensive apparatus.
“Our technology addresses these limitations by offering a real-time, remote, and cost-effective means of monitoring health conditions without the need for continuous oversight by healthcare professionals,” said research lead author and Izmir Institute of Technology Department of Bioengineering associate professor, Cumhur Tekin. “In the current era of healthcare’s digital transformation, we have harnessed the power of accessible and affordable hardware, coupled with advanced AI, to revolutionize the diagnosis and management of sleep apnea.”
Tekin explained that the wearable system is based on the open source electronic prototyping platform, Arduino, and in terms of what is worn by the patient, it consists of just an accelerometer placed on the patient’s diaphragm.
“This technology efficiently detects breathing patterns through the synergy of easily accessible sensors and AI-powered algorithms, all within a low-cost and ergonomic design,” he continued. “The developed wearable device offers several advantages to the patient. It is a cost-effective solution that eliminates the need for multiple devices or complex connections, such as those required by polysomnography.
“This allows patients to conveniently receive high-quality monitoring of their condition in the comfort of their own homes, without disrupting their sleep routine.”
AI watches you while you sleep
At the heart of the diagnosis procedure is an advanced image processing system that is based on deep learning. This enables the real time detection of various breathing patterns, including inhalation, exhalation, and instances of breathlessness.
To train the AI, the team simulated conditions resembling sleep apnea, inducing various breathing and breathless scenarios in healthy individuals across a range of different body positions.
“The accelerometer data of diaphragmatic movements collected from these individuals were then translated into a graphical representation,” said Tekin. “Our advanced image-processing system, enhanced by state-of-the-art You Look Only Once (YOLO) algorithms, effectively identifies respiratory patterns on this diaphragmatic movement graph.”
The system was able to reliably distinguish between inhalation and exhalation and could spot instances of breathlessness lasting longer than 10 seconds — key to diagnosing sleep apnea, according to Tekin. The team reported the system has an average of 96% accuracy in identifying breath patterns.
“Our advanced deep learning algorithms have significantly enhanced the accuracy of predicting breath patterns using a basic accelerometer,” Tekin continued.
Even the scientists behind this device were pleasantly surprised by its adaptability as it demonstrated its capacity to monitor breaths across various body positions accurately and even after significant body movements. Tekin added that this capability demonstrated the technology’s robustness and reliability in real-world scenarios.
“We were excited by the system’s real-time analysis capability, which opens up possibilities for its integration into various fields beyond sleep apnea diagnosis,” Tekin continued. “This includes applications in assessing other disorders and even in sports, where breath pattern analysis can be of significant value.”
A potential breakthrough
The research team hopes to further streamline and simplify the diagnostic process by promptly reporting any instances of breathlessness or abnormalities on the collected data during sleep to healthcare providers. This means that the system won’t just be beneficial to individual patients but to the healthcare industry as a whole.
“We have significantly expedited the detection process, enabling real-time recognition of breath patterns. This real-time capability enables remote monitoring of breath patterns, offering healthcare professionals valuable insights into a patient’s sleep breathing,” Tekin added. “This streamlined approach facilitates remote diagnosis and continuous monitoring of sleep apnea, ultimately alleviating the burden on hospitals and healthcare professionals.”
The system isn’t quite ready for rollout just yet, with the team now planning to integrate the device into a wireless framework, thus making it entirely independent and remotely controllable. They also aim to make the device more comfortable to wear during sleep by reducing its size.
“To bring this system to the mass market, we plan to validate its utility as a reliable tool for diagnosing and monitoring sleep apnea through extensive testing on patients with sleep apnea syndrome. This will position our technology within the sleep disorders market,” Tekin concluded.
“Additionally, we are committed to enhancing our system’s capabilities by developing improved algorithms that can differentiate between various types of sleep apnea disorders. This advancement will pave the way to applying more effective treatment strategies.”
Reference: H. C. Tekin., et al., A Wearable Device Integrated with Deep Learning-Based Algorithms for the Analysis of Breath Patterns, Advanced Intelligent Systems (2023). DOI:
Feature image credit: Quin Stevenson on Unsplash