Medical professionals at Onera Health have developed a bio-impedance patch that uses machine learning to help detect sleep apnea, a serious sleep disorder where breathing repeatedly starts and stops, which can lead to cardiovascular disease, memory problems, and other health-related issues. Diagnosing sleep apnea is done via an overnight study at special clinics, which use sensors to garner data on brain activity, eye movement, and blood oxygen levels. Onera’s patch could make it easier to identify sleep apnea in a wearer's own home, allowing more patients to be tested with long-term monitoring.
The bio-impedance patch, which is worn on the chest, applies a small current at a known frequency and measures the resulting voltage as it passes through the body at a different location. The readings are then analyzed using a two-phase LSTM (Long Short-Term Memory) deep learning algorithm to detect sleep apnea events. The bio-impedance patch is based on a device (known as Robin) from IMEC and Ghent University researchers, who wondered if it could be used to distinguish breathing patterns of those who have sleep apnea.
The researchers from Onera Health have trialed the device using simultaneous recordings and the polysomnography (records brain waves, breathing patterns, heart rate, etc.) of 25 patients, and found their system has a 73% accuracy in detecting the sleep disorder. Other equipment on the market use elastic bands or MEMS sensors to detect breathing patterns indirectly from the outside the body, but Onera’s does it internally within the body, making it more reliable. The researchers are currently working on ways to combine their deep learning sleep apnea unit with other physiological signal devices, as the American Academy of Sleep Medicine requires strict guidelines before a medical device can be approved for medical markets.