Sleep is an important part of our life considering that we spend almost a third of it doing it. A good night’s sleep is generally considered to be quite important for a refreshing and productive day ahead. However, unfortunately, the number of people reporting to have poor sleep quality is on the rise. It is estimated that more than 3.5 million people in the United Kingdom and more than 70 million people in United States suffer from some sort of sleep disorder. This has a huge financial impact on the economy stemming from expensive treatments, reduced productivity, road accidents and many other areas that involve alertness and quick judgement. Common sleep disorders include obstructive sleep apnea (OSA) causing disruption of sleep by interruption of breathing, REM Sleep Behaviour Disorder (RBD), narcolepsy, sleepwalking, night terrors, etc. Often, sleep disorders are not a problem on their own but a consequence of other conditions such as depression.
Diagnosis of sleep disorders is an expensive and time-consuming procedure that requires performing a sleep study, known as polysomnography, in a controlled environment. This normally involves using a large number of electrodes to monitor and record various physiological parameters together with brain activity, eye movements and muscle activity. These signals are then analysed and scored by clinicians to determine if there is any abnormality recorded during sleep.
At Wearable Technologies Lab, we are carrying out research aiming to take this whole procedure away from hospitals and bringing it at the patients’ homes. This not only saves costs but also provides a familiar and comfortable environment for the patients. We are developing tiny wearable devices that can be put on by patients easily during sleep which can monitor their neural activity, analyse the overnight signals, and automatically identify unusual patterns that can be helpful for their doctors.
To achieve this, we have developed several low complexity algorithms for automatic analysis and identification of different sleep abnormalities using information from only a limited number of channels with very high accuracy. Our algorithm development workflow involves strong focus on optimising all arithmetic operations, power optimisation as well as achieving high performance. Some of these algorithms have also been implemented using analogue and digital integrated circuits and consume only micro-watts of power. This allows the algorithms to be easily used at the sensor-end of a low power wearable device using very little processing resources. Combined with our tiny, lightweight, and low-power respiratory and brain monitoring devices, these algorithms enable real-time monitoring and analysis of sleep.
As part of our work we have also developed an Open Source toolbox to help other researchers getting started with the analysis of sleep EEG signals as well as standardise the way in which different sleep algorithms are evaluated.