Estimation of heart rate during sleep measured from a gyroscope embedded in a CPAP mask

Abstract

A device was developed to monitor the heart rate of patients who are undergoing CPAP therapy, a technique used to treat Sleep Apnea. This device is a wearable ballistocardiographic device that comprises of a 3-axis gyroscope, embedded into a full face CPAP mask. Data were collected from the gyroscope whilst being worn by a patient in simulated sleeping conditions. From these data, an algorithm was developed to extract heart rate signals from different heart rate estimates from the gyroscope. Over a 5 minute period, the algorithm was able to correctly identify between 64-95% of the heartbeats from the different gyroscope components. In order to improve the accuracy of the heart rate detection, two data fusion algorithms were implemented on the outputs from the heartrate detection algorithm. These two data fusion algorithms involved a custom weighted average algorithm and a Kalman filter, which when applied were able to achieve mean errors of 2.776 beats per minute (STD. 3.279) and 2.349 beats per minute (STD. 2.298) respectively. The performance of these algorithms in the presence of movement artefacts was also tested, and the tests showed that the algorithm is able to recover from large movement artefacts after a short period of between 7-15 seconds. These results show that it is possible to accurately monitor the HR of a subject using signals from a gyroscope embedded into a CPAP mask.